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>
47 #include <asm/byteorder.h>
48 #include <uapi/linux/fs.h>
50 struct backing_dev_info;
54 struct export_operations;
55 struct fiemap_extent_info;
60 struct pipe_inode_info;
61 struct poll_table_struct;
63 struct vm_area_struct;
66 struct swap_info_struct;
68 struct workqueue_struct;
71 struct fscrypt_operations;
73 struct fsverity_operations;
75 struct fs_parameter_spec;
79 extern void __init inode_init(void);
80 extern void __init inode_init_early(void);
81 extern void __init files_init(void);
82 extern void __init files_maxfiles_init(void);
84 extern unsigned long get_max_files(void);
85 extern unsigned int sysctl_nr_open;
87 typedef __kernel_rwf_t rwf_t;
90 typedef int (get_block_t)(struct inode *inode, sector_t iblock,
91 struct buffer_head *bh_result, int create);
92 typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
93 ssize_t bytes, void *private);
95 #define MAY_EXEC 0x00000001
96 #define MAY_WRITE 0x00000002
97 #define MAY_READ 0x00000004
98 #define MAY_APPEND 0x00000008
99 #define MAY_ACCESS 0x00000010
100 #define MAY_OPEN 0x00000020
101 #define MAY_CHDIR 0x00000040
102 /* called from RCU mode, don't block */
103 #define MAY_NOT_BLOCK 0x00000080
106 * flags in file.f_mode. Note that FMODE_READ and FMODE_WRITE must correspond
107 * to O_WRONLY and O_RDWR via the strange trick in do_dentry_open()
110 /* file is open for reading */
111 #define FMODE_READ ((__force fmode_t)0x1)
112 /* file is open for writing */
113 #define FMODE_WRITE ((__force fmode_t)0x2)
114 /* file is seekable */
115 #define FMODE_LSEEK ((__force fmode_t)0x4)
116 /* file can be accessed using pread */
117 #define FMODE_PREAD ((__force fmode_t)0x8)
118 /* file can be accessed using pwrite */
119 #define FMODE_PWRITE ((__force fmode_t)0x10)
120 /* File is opened for execution with sys_execve / sys_uselib */
121 #define FMODE_EXEC ((__force fmode_t)0x20)
122 /* 32bit hashes as llseek() offset (for directories) */
123 #define FMODE_32BITHASH ((__force fmode_t)0x200)
124 /* 64bit hashes as llseek() offset (for directories) */
125 #define FMODE_64BITHASH ((__force fmode_t)0x400)
128 * Don't update ctime and mtime.
130 * Currently a special hack for the XFS open_by_handle ioctl, but we'll
131 * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon.
133 #define FMODE_NOCMTIME ((__force fmode_t)0x800)
135 /* Expect random access pattern */
136 #define FMODE_RANDOM ((__force fmode_t)0x1000)
138 /* File is huge (eg. /dev/mem): treat loff_t as unsigned */
139 #define FMODE_UNSIGNED_OFFSET ((__force fmode_t)0x2000)
141 /* File is opened with O_PATH; almost nothing can be done with it */
142 #define FMODE_PATH ((__force fmode_t)0x4000)
144 /* File needs atomic accesses to f_pos */
145 #define FMODE_ATOMIC_POS ((__force fmode_t)0x8000)
146 /* Write access to underlying fs */
147 #define FMODE_WRITER ((__force fmode_t)0x10000)
148 /* Has read method(s) */
149 #define FMODE_CAN_READ ((__force fmode_t)0x20000)
150 /* Has write method(s) */
151 #define FMODE_CAN_WRITE ((__force fmode_t)0x40000)
153 #define FMODE_OPENED ((__force fmode_t)0x80000)
154 #define FMODE_CREATED ((__force fmode_t)0x100000)
156 /* File is stream-like */
157 #define FMODE_STREAM ((__force fmode_t)0x200000)
159 /* File supports DIRECT IO */
160 #define FMODE_CAN_ODIRECT ((__force fmode_t)0x400000)
162 #define FMODE_NOREUSE ((__force fmode_t)0x800000)
164 /* File supports non-exclusive O_DIRECT writes from multiple threads */
165 #define FMODE_DIO_PARALLEL_WRITE ((__force fmode_t)0x1000000)
167 /* File is embedded in backing_file object */
168 #define FMODE_BACKING ((__force fmode_t)0x2000000)
170 /* File was opened by fanotify and shouldn't generate fanotify events */
171 #define FMODE_NONOTIFY ((__force fmode_t)0x4000000)
173 /* File is capable of returning -EAGAIN if I/O will block */
174 #define FMODE_NOWAIT ((__force fmode_t)0x8000000)
176 /* File represents mount that needs unmounting */
177 #define FMODE_NEED_UNMOUNT ((__force fmode_t)0x10000000)
179 /* File does not contribute to nr_files count */
180 #define FMODE_NOACCOUNT ((__force fmode_t)0x20000000)
182 /* File supports async buffered reads */
183 #define FMODE_BUF_RASYNC ((__force fmode_t)0x40000000)
185 /* File supports async nowait buffered writes */
186 #define FMODE_BUF_WASYNC ((__force fmode_t)0x80000000)
189 * Attribute flags. These should be or-ed together to figure out what
192 #define ATTR_MODE (1 << 0)
193 #define ATTR_UID (1 << 1)
194 #define ATTR_GID (1 << 2)
195 #define ATTR_SIZE (1 << 3)
196 #define ATTR_ATIME (1 << 4)
197 #define ATTR_MTIME (1 << 5)
198 #define ATTR_CTIME (1 << 6)
199 #define ATTR_ATIME_SET (1 << 7)
200 #define ATTR_MTIME_SET (1 << 8)
201 #define ATTR_FORCE (1 << 9) /* Not a change, but a change it */
202 #define ATTR_KILL_SUID (1 << 11)
203 #define ATTR_KILL_SGID (1 << 12)
204 #define ATTR_FILE (1 << 13)
205 #define ATTR_KILL_PRIV (1 << 14)
206 #define ATTR_OPEN (1 << 15) /* Truncating from open(O_TRUNC) */
207 #define ATTR_TIMES_SET (1 << 16)
208 #define ATTR_TOUCH (1 << 17)
211 * Whiteout is represented by a char device. The following constants define the
212 * mode and device number to use.
214 #define WHITEOUT_MODE 0
215 #define WHITEOUT_DEV 0
218 * This is the Inode Attributes structure, used for notify_change(). It
219 * uses the above definitions as flags, to know which values have changed.
220 * Also, in this manner, a Filesystem can look at only the values it cares
221 * about. Basically, these are the attributes that the VFS layer can
222 * request to change from the FS layer.
224 * Derek Atkins <warlord@MIT.EDU> 94-10-20
227 unsigned int ia_valid;
230 * The two anonymous unions wrap structures with the same member.
232 * Filesystems raising FS_ALLOW_IDMAP need to use ia_vfs{g,u}id which
233 * are a dedicated type requiring the filesystem to use the dedicated
234 * helpers. Other filesystem can continue to use ia_{g,u}id until they
237 * They always contain the same value. In other words FS_ALLOW_IDMAP
238 * pass down the same value on idmapped mounts as they would on regular
250 struct timespec64 ia_atime;
251 struct timespec64 ia_mtime;
252 struct timespec64 ia_ctime;
255 * Not an attribute, but an auxiliary info for filesystems wanting to
256 * implement an ftruncate() like method. NOTE: filesystem should
257 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL).
259 struct file *ia_file;
263 * Includes for diskquotas.
265 #include <linux/quota.h>
268 * Maximum number of layers of fs stack. Needs to be limited to
269 * prevent kernel stack overflow
271 #define FILESYSTEM_MAX_STACK_DEPTH 2
274 * enum positive_aop_returns - aop return codes with specific semantics
276 * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has
277 * completed, that the page is still locked, and
278 * should be considered active. The VM uses this hint
279 * to return the page to the active list -- it won't
280 * be a candidate for writeback again in the near
281 * future. Other callers must be careful to unlock
282 * the page if they get this return. Returned by
285 * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has
286 * unlocked it and the page might have been truncated.
287 * The caller should back up to acquiring a new page and
288 * trying again. The aop will be taking reasonable
289 * precautions not to livelock. If the caller held a page
290 * reference, it should drop it before retrying. Returned
293 * address_space_operation functions return these large constants to indicate
294 * special semantics to the caller. These are much larger than the bytes in a
295 * page to allow for functions that return the number of bytes operated on in a
299 enum positive_aop_returns {
300 AOP_WRITEPAGE_ACTIVATE = 0x80000,
301 AOP_TRUNCATED_PAGE = 0x80001,
305 * oh the beauties of C type declarations.
308 struct address_space;
309 struct writeback_control;
310 struct readahead_control;
313 * Write life time hint values.
314 * Stored in struct inode as u8.
317 WRITE_LIFE_NOT_SET = 0,
318 WRITE_LIFE_NONE = RWH_WRITE_LIFE_NONE,
319 WRITE_LIFE_SHORT = RWH_WRITE_LIFE_SHORT,
320 WRITE_LIFE_MEDIUM = RWH_WRITE_LIFE_MEDIUM,
321 WRITE_LIFE_LONG = RWH_WRITE_LIFE_LONG,
322 WRITE_LIFE_EXTREME = RWH_WRITE_LIFE_EXTREME,
325 /* Match RWF_* bits to IOCB bits */
326 #define IOCB_HIPRI (__force int) RWF_HIPRI
327 #define IOCB_DSYNC (__force int) RWF_DSYNC
328 #define IOCB_SYNC (__force int) RWF_SYNC
329 #define IOCB_NOWAIT (__force int) RWF_NOWAIT
330 #define IOCB_APPEND (__force int) RWF_APPEND
332 /* non-RWF related bits - start at 16 */
333 #define IOCB_EVENTFD (1 << 16)
334 #define IOCB_DIRECT (1 << 17)
335 #define IOCB_WRITE (1 << 18)
336 /* iocb->ki_waitq is valid */
337 #define IOCB_WAITQ (1 << 19)
338 #define IOCB_NOIO (1 << 20)
339 /* can use bio alloc cache */
340 #define IOCB_ALLOC_CACHE (1 << 21)
342 /* for use in trace events */
343 #define TRACE_IOCB_STRINGS \
344 { IOCB_HIPRI, "HIPRI" }, \
345 { IOCB_DSYNC, "DSYNC" }, \
346 { IOCB_SYNC, "SYNC" }, \
347 { IOCB_NOWAIT, "NOWAIT" }, \
348 { IOCB_APPEND, "APPEND" }, \
349 { IOCB_EVENTFD, "EVENTFD"}, \
350 { IOCB_DIRECT, "DIRECT" }, \
351 { IOCB_WRITE, "WRITE" }, \
352 { IOCB_WAITQ, "WAITQ" }, \
353 { IOCB_NOIO, "NOIO" }, \
354 { IOCB_ALLOC_CACHE, "ALLOC_CACHE" }
357 struct file *ki_filp;
359 void (*ki_complete)(struct kiocb *iocb, long ret);
362 u16 ki_ioprio; /* See linux/ioprio.h */
363 struct wait_page_queue *ki_waitq; /* for async buffered IO */
366 static inline bool is_sync_kiocb(struct kiocb *kiocb)
368 return kiocb->ki_complete == NULL;
371 struct address_space_operations {
372 int (*writepage)(struct page *page, struct writeback_control *wbc);
373 int (*read_folio)(struct file *, struct folio *);
375 /* Write back some dirty pages from this mapping. */
376 int (*writepages)(struct address_space *, struct writeback_control *);
378 /* Mark a folio dirty. Return true if this dirtied it */
379 bool (*dirty_folio)(struct address_space *, struct folio *);
381 void (*readahead)(struct readahead_control *);
383 int (*write_begin)(struct file *, struct address_space *mapping,
384 loff_t pos, unsigned len,
385 struct page **pagep, void **fsdata);
386 int (*write_end)(struct file *, struct address_space *mapping,
387 loff_t pos, unsigned len, unsigned copied,
388 struct page *page, void *fsdata);
390 /* Unfortunately this kludge is needed for FIBMAP. Don't use it */
391 sector_t (*bmap)(struct address_space *, sector_t);
392 void (*invalidate_folio) (struct folio *, size_t offset, size_t len);
393 bool (*release_folio)(struct folio *, gfp_t);
394 void (*free_folio)(struct folio *folio);
395 ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter);
397 * migrate the contents of a folio to the specified target. If
398 * migrate_mode is MIGRATE_ASYNC, it must not block.
400 int (*migrate_folio)(struct address_space *, struct folio *dst,
401 struct folio *src, enum migrate_mode);
402 int (*launder_folio)(struct folio *);
403 bool (*is_partially_uptodate) (struct folio *, size_t from,
405 void (*is_dirty_writeback) (struct folio *, bool *dirty, bool *wb);
406 int (*error_remove_page)(struct address_space *, struct page *);
408 /* swapfile support */
409 int (*swap_activate)(struct swap_info_struct *sis, struct file *file,
411 void (*swap_deactivate)(struct file *file);
412 int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter);
415 extern const struct address_space_operations empty_aops;
418 * struct address_space - Contents of a cacheable, mappable object.
419 * @host: Owner, either the inode or the block_device.
420 * @i_pages: Cached pages.
421 * @invalidate_lock: Guards coherency between page cache contents and
422 * file offset->disk block mappings in the filesystem during invalidates.
423 * It is also used to block modification of page cache contents through
425 * @gfp_mask: Memory allocation flags to use for allocating pages.
426 * @i_mmap_writable: Number of VM_SHARED mappings.
427 * @nr_thps: Number of THPs in the pagecache (non-shmem only).
428 * @i_mmap: Tree of private and shared mappings.
429 * @i_mmap_rwsem: Protects @i_mmap and @i_mmap_writable.
430 * @nrpages: Number of page entries, protected by the i_pages lock.
431 * @writeback_index: Writeback starts here.
433 * @flags: Error bits and flags (AS_*).
434 * @wb_err: The most recent error which has occurred.
435 * @private_lock: For use by the owner of the address_space.
436 * @private_list: For use by the owner of the address_space.
437 * @private_data: For use by the owner of the address_space.
439 struct address_space {
441 struct xarray i_pages;
442 struct rw_semaphore invalidate_lock;
444 atomic_t i_mmap_writable;
445 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
446 /* number of thp, only for non-shmem files */
449 struct rb_root_cached i_mmap;
450 struct rw_semaphore i_mmap_rwsem;
451 unsigned long nrpages;
452 pgoff_t writeback_index;
453 const struct address_space_operations *a_ops;
456 spinlock_t private_lock;
457 struct list_head private_list;
459 } __attribute__((aligned(sizeof(long)))) __randomize_layout;
461 * On most architectures that alignment is already the case; but
462 * must be enforced here for CRIS, to let the least significant bit
463 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON.
466 /* XArray tags, for tagging dirty and writeback pages in the pagecache. */
467 #define PAGECACHE_TAG_DIRTY XA_MARK_0
468 #define PAGECACHE_TAG_WRITEBACK XA_MARK_1
469 #define PAGECACHE_TAG_TOWRITE XA_MARK_2
472 * Returns true if any of the pages in the mapping are marked with the tag.
474 static inline bool mapping_tagged(struct address_space *mapping, xa_mark_t tag)
476 return xa_marked(&mapping->i_pages, tag);
479 static inline void i_mmap_lock_write(struct address_space *mapping)
481 down_write(&mapping->i_mmap_rwsem);
484 static inline int i_mmap_trylock_write(struct address_space *mapping)
486 return down_write_trylock(&mapping->i_mmap_rwsem);
489 static inline void i_mmap_unlock_write(struct address_space *mapping)
491 up_write(&mapping->i_mmap_rwsem);
494 static inline int i_mmap_trylock_read(struct address_space *mapping)
496 return down_read_trylock(&mapping->i_mmap_rwsem);
499 static inline void i_mmap_lock_read(struct address_space *mapping)
501 down_read(&mapping->i_mmap_rwsem);
504 static inline void i_mmap_unlock_read(struct address_space *mapping)
506 up_read(&mapping->i_mmap_rwsem);
509 static inline void i_mmap_assert_locked(struct address_space *mapping)
511 lockdep_assert_held(&mapping->i_mmap_rwsem);
514 static inline void i_mmap_assert_write_locked(struct address_space *mapping)
516 lockdep_assert_held_write(&mapping->i_mmap_rwsem);
520 * Might pages of this file be mapped into userspace?
522 static inline int mapping_mapped(struct address_space *mapping)
524 return !RB_EMPTY_ROOT(&mapping->i_mmap.rb_root);
528 * Might pages of this file have been modified in userspace?
529 * Note that i_mmap_writable counts all VM_SHARED vmas: do_mmap
530 * marks vma as VM_SHARED if it is shared, and the file was opened for
531 * writing i.e. vma may be mprotected writable even if now readonly.
533 * If i_mmap_writable is negative, no new writable mappings are allowed. You
534 * can only deny writable mappings, if none exists right now.
536 static inline int mapping_writably_mapped(struct address_space *mapping)
538 return atomic_read(&mapping->i_mmap_writable) > 0;
541 static inline int mapping_map_writable(struct address_space *mapping)
543 return atomic_inc_unless_negative(&mapping->i_mmap_writable) ?
547 static inline void mapping_unmap_writable(struct address_space *mapping)
549 atomic_dec(&mapping->i_mmap_writable);
552 static inline int mapping_deny_writable(struct address_space *mapping)
554 return atomic_dec_unless_positive(&mapping->i_mmap_writable) ?
558 static inline void mapping_allow_writable(struct address_space *mapping)
560 atomic_inc(&mapping->i_mmap_writable);
564 * Use sequence counter to get consistent i_size on 32-bit processors.
566 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
567 #include <linux/seqlock.h>
568 #define __NEED_I_SIZE_ORDERED
569 #define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount)
571 #define i_size_ordered_init(inode) do { } while (0)
575 #define ACL_NOT_CACHED ((void *)(-1))
577 * ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to
578 * cache the ACL. This also means that ->get_inode_acl() can be called in RCU
579 * mode with the LOOKUP_RCU flag.
581 #define ACL_DONT_CACHE ((void *)(-3))
583 static inline struct posix_acl *
584 uncached_acl_sentinel(struct task_struct *task)
586 return (void *)task + 1;
590 is_uncached_acl(struct posix_acl *acl)
592 return (long)acl & 1;
595 #define IOP_FASTPERM 0x0001
596 #define IOP_LOOKUP 0x0002
597 #define IOP_NOFOLLOW 0x0004
598 #define IOP_XATTR 0x0008
599 #define IOP_DEFAULT_READLINK 0x0010
601 struct fsnotify_mark_connector;
604 * Keep mostly read-only and often accessed (especially for
605 * the RCU path lookup and 'stat' data) fields at the beginning
606 * of the 'struct inode'
610 unsigned short i_opflags;
613 unsigned int i_flags;
615 #ifdef CONFIG_FS_POSIX_ACL
616 struct posix_acl *i_acl;
617 struct posix_acl *i_default_acl;
620 const struct inode_operations *i_op;
621 struct super_block *i_sb;
622 struct address_space *i_mapping;
624 #ifdef CONFIG_SECURITY
628 /* Stat data, not accessed from path walking */
631 * Filesystems may only read i_nlink directly. They shall use the
632 * following functions for modification:
634 * (set|clear|inc|drop)_nlink
635 * inode_(inc|dec)_link_count
638 const unsigned int i_nlink;
639 unsigned int __i_nlink;
643 struct timespec64 i_atime;
644 struct timespec64 i_mtime;
645 struct timespec64 i_ctime;
646 spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */
647 unsigned short i_bytes;
652 #ifdef __NEED_I_SIZE_ORDERED
653 seqcount_t i_size_seqcount;
657 unsigned long i_state;
658 struct rw_semaphore i_rwsem;
660 unsigned long dirtied_when; /* jiffies of first dirtying */
661 unsigned long dirtied_time_when;
663 struct hlist_node i_hash;
664 struct list_head i_io_list; /* backing dev IO list */
665 #ifdef CONFIG_CGROUP_WRITEBACK
666 struct bdi_writeback *i_wb; /* the associated cgroup wb */
668 /* foreign inode detection, see wbc_detach_inode() */
670 u16 i_wb_frn_avg_time;
671 u16 i_wb_frn_history;
673 struct list_head i_lru; /* inode LRU list */
674 struct list_head i_sb_list;
675 struct list_head i_wb_list; /* backing dev writeback list */
677 struct hlist_head i_dentry;
678 struct rcu_head i_rcu;
680 atomic64_t i_version;
681 atomic64_t i_sequence; /* see futex */
683 atomic_t i_dio_count;
684 atomic_t i_writecount;
685 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
686 atomic_t i_readcount; /* struct files open RO */
689 const struct file_operations *i_fop; /* former ->i_op->default_file_ops */
690 void (*free_inode)(struct inode *);
692 struct file_lock_context *i_flctx;
693 struct address_space i_data;
694 struct list_head i_devices;
696 struct pipe_inode_info *i_pipe;
704 #ifdef CONFIG_FSNOTIFY
705 __u32 i_fsnotify_mask; /* all events this inode cares about */
706 struct fsnotify_mark_connector __rcu *i_fsnotify_marks;
709 #ifdef CONFIG_FS_ENCRYPTION
710 struct fscrypt_info *i_crypt_info;
713 #ifdef CONFIG_FS_VERITY
714 struct fsverity_info *i_verity_info;
717 void *i_private; /* fs or device private pointer */
718 } __randomize_layout;
720 struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode);
722 static inline unsigned int i_blocksize(const struct inode *node)
724 return (1 << node->i_blkbits);
727 static inline int inode_unhashed(struct inode *inode)
729 return hlist_unhashed(&inode->i_hash);
733 * __mark_inode_dirty expects inodes to be hashed. Since we don't
734 * want special inodes in the fileset inode space, we make them
735 * appear hashed, but do not put on any lists. hlist_del()
736 * will work fine and require no locking.
738 static inline void inode_fake_hash(struct inode *inode)
740 hlist_add_fake(&inode->i_hash);
744 * inode->i_mutex nesting subclasses for the lock validator:
746 * 0: the object of the current VFS operation
750 * 4: second non-directory
751 * 5: second parent (when locking independent directories in rename)
753 * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two
754 * non-directories at once.
756 * The locking order between these classes is
757 * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory
759 enum inode_i_mutex_lock_class
769 static inline void inode_lock(struct inode *inode)
771 down_write(&inode->i_rwsem);
774 static inline void inode_unlock(struct inode *inode)
776 up_write(&inode->i_rwsem);
779 static inline void inode_lock_shared(struct inode *inode)
781 down_read(&inode->i_rwsem);
784 static inline void inode_unlock_shared(struct inode *inode)
786 up_read(&inode->i_rwsem);
789 static inline int inode_trylock(struct inode *inode)
791 return down_write_trylock(&inode->i_rwsem);
794 static inline int inode_trylock_shared(struct inode *inode)
796 return down_read_trylock(&inode->i_rwsem);
799 static inline int inode_is_locked(struct inode *inode)
801 return rwsem_is_locked(&inode->i_rwsem);
804 static inline void inode_lock_nested(struct inode *inode, unsigned subclass)
806 down_write_nested(&inode->i_rwsem, subclass);
809 static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass)
811 down_read_nested(&inode->i_rwsem, subclass);
814 static inline void filemap_invalidate_lock(struct address_space *mapping)
816 down_write(&mapping->invalidate_lock);
819 static inline void filemap_invalidate_unlock(struct address_space *mapping)
821 up_write(&mapping->invalidate_lock);
824 static inline void filemap_invalidate_lock_shared(struct address_space *mapping)
826 down_read(&mapping->invalidate_lock);
829 static inline int filemap_invalidate_trylock_shared(
830 struct address_space *mapping)
832 return down_read_trylock(&mapping->invalidate_lock);
835 static inline void filemap_invalidate_unlock_shared(
836 struct address_space *mapping)
838 up_read(&mapping->invalidate_lock);
841 void lock_two_nondirectories(struct inode *, struct inode*);
842 void unlock_two_nondirectories(struct inode *, struct inode*);
844 void filemap_invalidate_lock_two(struct address_space *mapping1,
845 struct address_space *mapping2);
846 void filemap_invalidate_unlock_two(struct address_space *mapping1,
847 struct address_space *mapping2);
851 * NOTE: in a 32bit arch with a preemptable kernel and
852 * an UP compile the i_size_read/write must be atomic
853 * with respect to the local cpu (unlike with preempt disabled),
854 * but they don't need to be atomic with respect to other cpus like in
855 * true SMP (so they need either to either locally disable irq around
856 * the read or for example on x86 they can be still implemented as a
857 * cmpxchg8b without the need of the lock prefix). For SMP compiles
858 * and 64bit archs it makes no difference if preempt is enabled or not.
860 static inline loff_t i_size_read(const struct inode *inode)
862 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
867 seq = read_seqcount_begin(&inode->i_size_seqcount);
868 i_size = inode->i_size;
869 } while (read_seqcount_retry(&inode->i_size_seqcount, seq));
871 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
875 i_size = inode->i_size;
879 return inode->i_size;
884 * NOTE: unlike i_size_read(), i_size_write() does need locking around it
885 * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount
886 * can be lost, resulting in subsequent i_size_read() calls spinning forever.
888 static inline void i_size_write(struct inode *inode, loff_t i_size)
890 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
892 write_seqcount_begin(&inode->i_size_seqcount);
893 inode->i_size = i_size;
894 write_seqcount_end(&inode->i_size_seqcount);
896 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
898 inode->i_size = i_size;
901 inode->i_size = i_size;
905 static inline unsigned iminor(const struct inode *inode)
907 return MINOR(inode->i_rdev);
910 static inline unsigned imajor(const struct inode *inode)
912 return MAJOR(inode->i_rdev);
916 rwlock_t lock; /* protects pid, uid, euid fields */
917 struct pid *pid; /* pid or -pgrp where SIGIO should be sent */
918 enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */
919 kuid_t uid, euid; /* uid/euid of process setting the owner */
920 int signum; /* posix.1b rt signal to be delivered on IO */
924 * struct file_ra_state - Track a file's readahead state.
925 * @start: Where the most recent readahead started.
926 * @size: Number of pages read in the most recent readahead.
927 * @async_size: Numer of pages that were/are not needed immediately
928 * and so were/are genuinely "ahead". Start next readahead when
929 * the first of these pages is accessed.
930 * @ra_pages: Maximum size of a readahead request, copied from the bdi.
931 * @mmap_miss: How many mmap accesses missed in the page cache.
932 * @prev_pos: The last byte in the most recent read request.
934 * When this structure is passed to ->readahead(), the "most recent"
935 * readahead means the current readahead.
937 struct file_ra_state {
940 unsigned int async_size;
941 unsigned int ra_pages;
942 unsigned int mmap_miss;
947 * Check if @index falls in the readahead windows.
949 static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index)
951 return (index >= ra->start &&
952 index < ra->start + ra->size);
956 * f_{lock,count,pos_lock} members can be highly contended and share
957 * the same cacheline. f_{lock,mode} are very frequently used together
958 * and so share the same cacheline as well. The read-mostly
959 * f_{path,inode,op} are kept on a separate cacheline.
963 struct llist_node f_llist;
964 struct rcu_head f_rcuhead;
965 unsigned int f_iocb_flags;
969 * Protects f_ep, f_flags.
970 * Must not be taken from IRQ context.
974 atomic_long_t f_count;
975 struct mutex f_pos_lock;
977 unsigned int f_flags;
978 struct fown_struct f_owner;
979 const struct cred *f_cred;
980 struct file_ra_state f_ra;
982 struct inode *f_inode; /* cached value */
983 const struct file_operations *f_op;
986 #ifdef CONFIG_SECURITY
989 /* needed for tty driver, and maybe others */
993 /* Used by fs/eventpoll.c to link all the hooks to this file */
994 struct hlist_head *f_ep;
995 #endif /* #ifdef CONFIG_EPOLL */
996 struct address_space *f_mapping;
998 errseq_t f_sb_err; /* for syncfs */
1000 __attribute__((aligned(4))); /* lest something weird decides that 2 is OK */
1002 struct file_handle {
1005 /* file identifier */
1006 unsigned char f_handle[];
1009 static inline struct file *get_file(struct file *f)
1011 atomic_long_inc(&f->f_count);
1014 #define get_file_rcu(x) atomic_long_inc_not_zero(&(x)->f_count)
1015 #define file_count(x) atomic_long_read(&(x)->f_count)
1017 #define MAX_NON_LFS ((1UL<<31) - 1)
1019 /* Page cache limit. The filesystems should put that into their s_maxbytes
1020 limits, otherwise bad things can happen in VM. */
1021 #if BITS_PER_LONG==32
1022 #define MAX_LFS_FILESIZE ((loff_t)ULONG_MAX << PAGE_SHIFT)
1023 #elif BITS_PER_LONG==64
1024 #define MAX_LFS_FILESIZE ((loff_t)LLONG_MAX)
1027 /* legacy typedef, should eventually be removed */
1028 typedef void *fl_owner_t;
1032 /* The following constant reflects the upper bound of the file/locking space */
1034 #define OFFSET_MAX type_max(loff_t)
1035 #define OFFT_OFFSET_MAX type_max(off_t)
1038 extern void send_sigio(struct fown_struct *fown, int fd, int band);
1040 static inline struct inode *file_inode(const struct file *f)
1045 static inline struct dentry *file_dentry(const struct file *file)
1047 return d_real(file->f_path.dentry, file_inode(file));
1050 struct fasync_struct {
1054 struct fasync_struct *fa_next; /* singly linked list */
1055 struct file *fa_file;
1056 struct rcu_head fa_rcu;
1059 #define FASYNC_MAGIC 0x4601
1061 /* SMP safe fasync helpers: */
1062 extern int fasync_helper(int, struct file *, int, struct fasync_struct **);
1063 extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *);
1064 extern int fasync_remove_entry(struct file *, struct fasync_struct **);
1065 extern struct fasync_struct *fasync_alloc(void);
1066 extern void fasync_free(struct fasync_struct *);
1068 /* can be called from interrupts */
1069 extern void kill_fasync(struct fasync_struct **, int, int);
1071 extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force);
1072 extern int f_setown(struct file *filp, unsigned long arg, int force);
1073 extern void f_delown(struct file *filp);
1074 extern pid_t f_getown(struct file *filp);
1075 extern int send_sigurg(struct fown_struct *fown);
1078 * sb->s_flags. Note that these mirror the equivalent MS_* flags where
1079 * represented in both.
1081 #define SB_RDONLY BIT(0) /* Mount read-only */
1082 #define SB_NOSUID BIT(1) /* Ignore suid and sgid bits */
1083 #define SB_NODEV BIT(2) /* Disallow access to device special files */
1084 #define SB_NOEXEC BIT(3) /* Disallow program execution */
1085 #define SB_SYNCHRONOUS BIT(4) /* Writes are synced at once */
1086 #define SB_MANDLOCK BIT(6) /* Allow mandatory locks on an FS */
1087 #define SB_DIRSYNC BIT(7) /* Directory modifications are synchronous */
1088 #define SB_NOATIME BIT(10) /* Do not update access times. */
1089 #define SB_NODIRATIME BIT(11) /* Do not update directory access times */
1090 #define SB_SILENT BIT(15)
1091 #define SB_POSIXACL BIT(16) /* VFS does not apply the umask */
1092 #define SB_INLINECRYPT BIT(17) /* Use blk-crypto for encrypted files */
1093 #define SB_KERNMOUNT BIT(22) /* this is a kern_mount call */
1094 #define SB_I_VERSION BIT(23) /* Update inode I_version field */
1095 #define SB_LAZYTIME BIT(25) /* Update the on-disk [acm]times lazily */
1097 /* These sb flags are internal to the kernel */
1098 #define SB_SUBMOUNT BIT(26)
1099 #define SB_FORCE BIT(27)
1100 #define SB_NOSEC BIT(28)
1101 #define SB_BORN BIT(29)
1102 #define SB_ACTIVE BIT(30)
1103 #define SB_NOUSER BIT(31)
1105 /* These flags relate to encoding and casefolding */
1106 #define SB_ENC_STRICT_MODE_FL (1 << 0)
1108 #define sb_has_strict_encoding(sb) \
1109 (sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL)
1115 #define MNT_FORCE 0x00000001 /* Attempt to forcibily umount */
1116 #define MNT_DETACH 0x00000002 /* Just detach from the tree */
1117 #define MNT_EXPIRE 0x00000004 /* Mark for expiry */
1118 #define UMOUNT_NOFOLLOW 0x00000008 /* Don't follow symlink on umount */
1119 #define UMOUNT_UNUSED 0x80000000 /* Flag guaranteed to be unused */
1122 #define SB_I_CGROUPWB 0x00000001 /* cgroup-aware writeback enabled */
1123 #define SB_I_NOEXEC 0x00000002 /* Ignore executables on this fs */
1124 #define SB_I_NODEV 0x00000004 /* Ignore devices on this fs */
1125 #define SB_I_STABLE_WRITES 0x00000008 /* don't modify blks until WB is done */
1127 /* sb->s_iflags to limit user namespace mounts */
1128 #define SB_I_USERNS_VISIBLE 0x00000010 /* fstype already mounted */
1129 #define SB_I_IMA_UNVERIFIABLE_SIGNATURE 0x00000020
1130 #define SB_I_UNTRUSTED_MOUNTER 0x00000040
1132 #define SB_I_SKIP_SYNC 0x00000100 /* Skip superblock at global sync */
1133 #define SB_I_PERSB_BDI 0x00000200 /* has a per-sb bdi */
1134 #define SB_I_TS_EXPIRY_WARNED 0x00000400 /* warned about timestamp range expiry */
1135 #define SB_I_RETIRED 0x00000800 /* superblock shouldn't be reused */
1137 /* Possible states of 'frozen' field */
1139 SB_UNFROZEN = 0, /* FS is unfrozen */
1140 SB_FREEZE_WRITE = 1, /* Writes, dir ops, ioctls frozen */
1141 SB_FREEZE_PAGEFAULT = 2, /* Page faults stopped as well */
1142 SB_FREEZE_FS = 3, /* For internal FS use (e.g. to stop
1143 * internal threads if needed) */
1144 SB_FREEZE_COMPLETE = 4, /* ->freeze_fs finished successfully */
1147 #define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1)
1150 int frozen; /* Is sb frozen? */
1151 wait_queue_head_t wait_unfrozen; /* wait for thaw */
1152 struct percpu_rw_semaphore rw_sem[SB_FREEZE_LEVELS];
1155 struct super_block {
1156 struct list_head s_list; /* Keep this first */
1157 dev_t s_dev; /* search index; _not_ kdev_t */
1158 unsigned char s_blocksize_bits;
1159 unsigned long s_blocksize;
1160 loff_t s_maxbytes; /* Max file size */
1161 struct file_system_type *s_type;
1162 const struct super_operations *s_op;
1163 const struct dquot_operations *dq_op;
1164 const struct quotactl_ops *s_qcop;
1165 const struct export_operations *s_export_op;
1166 unsigned long s_flags;
1167 unsigned long s_iflags; /* internal SB_I_* flags */
1168 unsigned long s_magic;
1169 struct dentry *s_root;
1170 struct rw_semaphore s_umount;
1173 #ifdef CONFIG_SECURITY
1176 const struct xattr_handler **s_xattr;
1177 #ifdef CONFIG_FS_ENCRYPTION
1178 const struct fscrypt_operations *s_cop;
1179 struct fscrypt_keyring *s_master_keys; /* master crypto keys in use */
1181 #ifdef CONFIG_FS_VERITY
1182 const struct fsverity_operations *s_vop;
1184 #if IS_ENABLED(CONFIG_UNICODE)
1185 struct unicode_map *s_encoding;
1186 __u16 s_encoding_flags;
1188 struct hlist_bl_head s_roots; /* alternate root dentries for NFS */
1189 struct list_head s_mounts; /* list of mounts; _not_ for fs use */
1190 struct block_device *s_bdev;
1191 struct backing_dev_info *s_bdi;
1192 struct mtd_info *s_mtd;
1193 struct hlist_node s_instances;
1194 unsigned int s_quota_types; /* Bitmask of supported quota types */
1195 struct quota_info s_dquot; /* Diskquota specific options */
1197 struct sb_writers s_writers;
1200 * Keep s_fs_info, s_time_gran, s_fsnotify_mask, and
1201 * s_fsnotify_marks together for cache efficiency. They are frequently
1202 * accessed and rarely modified.
1204 void *s_fs_info; /* Filesystem private info */
1206 /* Granularity of c/m/atime in ns (cannot be worse than a second) */
1208 /* Time limits for c/m/atime in seconds */
1209 time64_t s_time_min;
1210 time64_t s_time_max;
1211 #ifdef CONFIG_FSNOTIFY
1212 __u32 s_fsnotify_mask;
1213 struct fsnotify_mark_connector __rcu *s_fsnotify_marks;
1216 char s_id[32]; /* Informational name */
1217 uuid_t s_uuid; /* UUID */
1219 unsigned int s_max_links;
1222 * The next field is for VFS *only*. No filesystems have any business
1223 * even looking at it. You had been warned.
1225 struct mutex s_vfs_rename_mutex; /* Kludge */
1228 * Filesystem subtype. If non-empty the filesystem type field
1229 * in /proc/mounts will be "type.subtype"
1231 const char *s_subtype;
1233 const struct dentry_operations *s_d_op; /* default d_op for dentries */
1235 struct shrinker s_shrink; /* per-sb shrinker handle */
1237 /* Number of inodes with nlink == 0 but still referenced */
1238 atomic_long_t s_remove_count;
1241 * Number of inode/mount/sb objects that are being watched, note that
1242 * inodes objects are currently double-accounted.
1244 atomic_long_t s_fsnotify_connectors;
1246 /* Read-only state of the superblock is being changed */
1247 int s_readonly_remount;
1249 /* per-sb errseq_t for reporting writeback errors via syncfs */
1252 /* AIO completions deferred from interrupt context */
1253 struct workqueue_struct *s_dio_done_wq;
1254 struct hlist_head s_pins;
1257 * Owning user namespace and default context in which to
1258 * interpret filesystem uids, gids, quotas, device nodes,
1259 * xattrs and security labels.
1261 struct user_namespace *s_user_ns;
1264 * The list_lru structure is essentially just a pointer to a table
1265 * of per-node lru lists, each of which has its own spinlock.
1266 * There is no need to put them into separate cachelines.
1268 struct list_lru s_dentry_lru;
1269 struct list_lru s_inode_lru;
1270 struct rcu_head rcu;
1271 struct work_struct destroy_work;
1273 struct mutex s_sync_lock; /* sync serialisation lock */
1276 * Indicates how deep in a filesystem stack this SB is
1280 /* s_inode_list_lock protects s_inodes */
1281 spinlock_t s_inode_list_lock ____cacheline_aligned_in_smp;
1282 struct list_head s_inodes; /* all inodes */
1284 spinlock_t s_inode_wblist_lock;
1285 struct list_head s_inodes_wb; /* writeback inodes */
1286 } __randomize_layout;
1288 static inline struct user_namespace *i_user_ns(const struct inode *inode)
1290 return inode->i_sb->s_user_ns;
1293 /* Helper functions so that in most cases filesystems will
1294 * not need to deal directly with kuid_t and kgid_t and can
1295 * instead deal with the raw numeric values that are stored
1296 * in the filesystem.
1298 static inline uid_t i_uid_read(const struct inode *inode)
1300 return from_kuid(i_user_ns(inode), inode->i_uid);
1303 static inline gid_t i_gid_read(const struct inode *inode)
1305 return from_kgid(i_user_ns(inode), inode->i_gid);
1308 static inline void i_uid_write(struct inode *inode, uid_t uid)
1310 inode->i_uid = make_kuid(i_user_ns(inode), uid);
1313 static inline void i_gid_write(struct inode *inode, gid_t gid)
1315 inode->i_gid = make_kgid(i_user_ns(inode), gid);
1319 * i_uid_into_vfsuid - map an inode's i_uid down according to an idmapping
1320 * @idmap: idmap of the mount the inode was found from
1321 * @inode: inode to map
1323 * Return: whe inode's i_uid mapped down according to @idmap.
1324 * If the inode's i_uid has no mapping INVALID_VFSUID is returned.
1326 static inline vfsuid_t i_uid_into_vfsuid(struct mnt_idmap *idmap,
1327 const struct inode *inode)
1329 return make_vfsuid(idmap, i_user_ns(inode), inode->i_uid);
1333 * i_uid_needs_update - check whether inode's i_uid needs to be updated
1334 * @idmap: idmap of the mount the inode was found from
1335 * @attr: the new attributes of @inode
1336 * @inode: the inode to update
1338 * Check whether the $inode's i_uid field needs to be updated taking idmapped
1339 * mounts into account if the filesystem supports it.
1341 * Return: true if @inode's i_uid field needs to be updated, false if not.
1343 static inline bool i_uid_needs_update(struct mnt_idmap *idmap,
1344 const struct iattr *attr,
1345 const struct inode *inode)
1347 return ((attr->ia_valid & ATTR_UID) &&
1348 !vfsuid_eq(attr->ia_vfsuid,
1349 i_uid_into_vfsuid(idmap, inode)));
1353 * i_uid_update - update @inode's i_uid field
1354 * @idmap: idmap of the mount the inode was found from
1355 * @attr: the new attributes of @inode
1356 * @inode: the inode to update
1358 * Safely update @inode's i_uid field translating the vfsuid of any idmapped
1359 * mount into the filesystem kuid.
1361 static inline void i_uid_update(struct mnt_idmap *idmap,
1362 const struct iattr *attr,
1363 struct inode *inode)
1365 if (attr->ia_valid & ATTR_UID)
1366 inode->i_uid = from_vfsuid(idmap, i_user_ns(inode),
1371 * i_gid_into_vfsgid - map an inode's i_gid down according to an idmapping
1372 * @idmap: idmap of the mount the inode was found from
1373 * @inode: inode to map
1375 * Return: the inode's i_gid mapped down according to @idmap.
1376 * If the inode's i_gid has no mapping INVALID_VFSGID is returned.
1378 static inline vfsgid_t i_gid_into_vfsgid(struct mnt_idmap *idmap,
1379 const struct inode *inode)
1381 return make_vfsgid(idmap, i_user_ns(inode), inode->i_gid);
1385 * i_gid_needs_update - check whether inode's i_gid needs to be updated
1386 * @idmap: idmap of the mount the inode was found from
1387 * @attr: the new attributes of @inode
1388 * @inode: the inode to update
1390 * Check whether the $inode's i_gid field needs to be updated taking idmapped
1391 * mounts into account if the filesystem supports it.
1393 * Return: true if @inode's i_gid field needs to be updated, false if not.
1395 static inline bool i_gid_needs_update(struct mnt_idmap *idmap,
1396 const struct iattr *attr,
1397 const struct inode *inode)
1399 return ((attr->ia_valid & ATTR_GID) &&
1400 !vfsgid_eq(attr->ia_vfsgid,
1401 i_gid_into_vfsgid(idmap, inode)));
1405 * i_gid_update - update @inode's i_gid field
1406 * @idmap: idmap of the mount the inode was found from
1407 * @attr: the new attributes of @inode
1408 * @inode: the inode to update
1410 * Safely update @inode's i_gid field translating the vfsgid of any idmapped
1411 * mount into the filesystem kgid.
1413 static inline void i_gid_update(struct mnt_idmap *idmap,
1414 const struct iattr *attr,
1415 struct inode *inode)
1417 if (attr->ia_valid & ATTR_GID)
1418 inode->i_gid = from_vfsgid(idmap, i_user_ns(inode),
1423 * inode_fsuid_set - initialize inode's i_uid field with callers fsuid
1424 * @inode: inode to initialize
1425 * @idmap: idmap of the mount the inode was found from
1427 * Initialize the i_uid field of @inode. If the inode was found/created via
1428 * an idmapped mount map the caller's fsuid according to @idmap.
1430 static inline void inode_fsuid_set(struct inode *inode,
1431 struct mnt_idmap *idmap)
1433 inode->i_uid = mapped_fsuid(idmap, i_user_ns(inode));
1437 * inode_fsgid_set - initialize inode's i_gid field with callers fsgid
1438 * @inode: inode to initialize
1439 * @idmap: idmap of the mount the inode was found from
1441 * Initialize the i_gid field of @inode. If the inode was found/created via
1442 * an idmapped mount map the caller's fsgid according to @idmap.
1444 static inline void inode_fsgid_set(struct inode *inode,
1445 struct mnt_idmap *idmap)
1447 inode->i_gid = mapped_fsgid(idmap, i_user_ns(inode));
1451 * fsuidgid_has_mapping() - check whether caller's fsuid/fsgid is mapped
1452 * @sb: the superblock we want a mapping in
1453 * @idmap: idmap of the relevant mount
1455 * Check whether the caller's fsuid and fsgid have a valid mapping in the
1456 * s_user_ns of the superblock @sb. If the caller is on an idmapped mount map
1457 * the caller's fsuid and fsgid according to the @idmap first.
1459 * Return: true if fsuid and fsgid is mapped, false if not.
1461 static inline bool fsuidgid_has_mapping(struct super_block *sb,
1462 struct mnt_idmap *idmap)
1464 struct user_namespace *fs_userns = sb->s_user_ns;
1468 kuid = mapped_fsuid(idmap, fs_userns);
1469 if (!uid_valid(kuid))
1471 kgid = mapped_fsgid(idmap, fs_userns);
1472 if (!gid_valid(kgid))
1474 return kuid_has_mapping(fs_userns, kuid) &&
1475 kgid_has_mapping(fs_userns, kgid);
1478 extern struct timespec64 current_time(struct inode *inode);
1481 * Snapshotting support.
1485 * These are internal functions, please use sb_start_{write,pagefault,intwrite}
1488 static inline void __sb_end_write(struct super_block *sb, int level)
1490 percpu_up_read(sb->s_writers.rw_sem + level-1);
1493 static inline void __sb_start_write(struct super_block *sb, int level)
1495 percpu_down_read(sb->s_writers.rw_sem + level - 1);
1498 static inline bool __sb_start_write_trylock(struct super_block *sb, int level)
1500 return percpu_down_read_trylock(sb->s_writers.rw_sem + level - 1);
1503 #define __sb_writers_acquired(sb, lev) \
1504 percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1505 #define __sb_writers_release(sb, lev) \
1506 percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1508 static inline bool sb_write_started(const struct super_block *sb)
1510 return lockdep_is_held_type(sb->s_writers.rw_sem + SB_FREEZE_WRITE - 1, 1);
1514 * sb_end_write - drop write access to a superblock
1515 * @sb: the super we wrote to
1517 * Decrement number of writers to the filesystem. Wake up possible waiters
1518 * wanting to freeze the filesystem.
1520 static inline void sb_end_write(struct super_block *sb)
1522 __sb_end_write(sb, SB_FREEZE_WRITE);
1526 * sb_end_pagefault - drop write access to a superblock from a page fault
1527 * @sb: the super we wrote to
1529 * Decrement number of processes handling write page fault to the filesystem.
1530 * Wake up possible waiters wanting to freeze the filesystem.
1532 static inline void sb_end_pagefault(struct super_block *sb)
1534 __sb_end_write(sb, SB_FREEZE_PAGEFAULT);
1538 * sb_end_intwrite - drop write access to a superblock for internal fs purposes
1539 * @sb: the super we wrote to
1541 * Decrement fs-internal number of writers to the filesystem. Wake up possible
1542 * waiters wanting to freeze the filesystem.
1544 static inline void sb_end_intwrite(struct super_block *sb)
1546 __sb_end_write(sb, SB_FREEZE_FS);
1550 * sb_start_write - get write access to a superblock
1551 * @sb: the super we write to
1553 * When a process wants to write data or metadata to a file system (i.e. dirty
1554 * a page or an inode), it should embed the operation in a sb_start_write() -
1555 * sb_end_write() pair to get exclusion against file system freezing. This
1556 * function increments number of writers preventing freezing. If the file
1557 * system is already frozen, the function waits until the file system is
1560 * Since freeze protection behaves as a lock, users have to preserve
1561 * ordering of freeze protection and other filesystem locks. Generally,
1562 * freeze protection should be the outermost lock. In particular, we have:
1565 * -> i_mutex (write path, truncate, directory ops, ...)
1566 * -> s_umount (freeze_super, thaw_super)
1568 static inline void sb_start_write(struct super_block *sb)
1570 __sb_start_write(sb, SB_FREEZE_WRITE);
1573 static inline bool sb_start_write_trylock(struct super_block *sb)
1575 return __sb_start_write_trylock(sb, SB_FREEZE_WRITE);
1579 * sb_start_pagefault - get write access to a superblock from a page fault
1580 * @sb: the super we write to
1582 * When a process starts handling write page fault, it should embed the
1583 * operation into sb_start_pagefault() - sb_end_pagefault() pair to get
1584 * exclusion against file system freezing. This is needed since the page fault
1585 * is going to dirty a page. This function increments number of running page
1586 * faults preventing freezing. If the file system is already frozen, the
1587 * function waits until the file system is thawed.
1589 * Since page fault freeze protection behaves as a lock, users have to preserve
1590 * ordering of freeze protection and other filesystem locks. It is advised to
1591 * put sb_start_pagefault() close to mmap_lock in lock ordering. Page fault
1592 * handling code implies lock dependency:
1595 * -> sb_start_pagefault
1597 static inline void sb_start_pagefault(struct super_block *sb)
1599 __sb_start_write(sb, SB_FREEZE_PAGEFAULT);
1603 * sb_start_intwrite - get write access to a superblock for internal fs purposes
1604 * @sb: the super we write to
1606 * This is the third level of protection against filesystem freezing. It is
1607 * free for use by a filesystem. The only requirement is that it must rank
1608 * below sb_start_pagefault.
1610 * For example filesystem can call sb_start_intwrite() when starting a
1611 * transaction which somewhat eases handling of freezing for internal sources
1612 * of filesystem changes (internal fs threads, discarding preallocation on file
1615 static inline void sb_start_intwrite(struct super_block *sb)
1617 __sb_start_write(sb, SB_FREEZE_FS);
1620 static inline bool sb_start_intwrite_trylock(struct super_block *sb)
1622 return __sb_start_write_trylock(sb, SB_FREEZE_FS);
1625 bool inode_owner_or_capable(struct mnt_idmap *idmap,
1626 const struct inode *inode);
1629 * VFS helper functions..
1631 int vfs_create(struct mnt_idmap *, struct inode *,
1632 struct dentry *, umode_t, bool);
1633 int vfs_mkdir(struct mnt_idmap *, struct inode *,
1634 struct dentry *, umode_t);
1635 int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *,
1637 int vfs_symlink(struct mnt_idmap *, struct inode *,
1638 struct dentry *, const char *);
1639 int vfs_link(struct dentry *, struct mnt_idmap *, struct inode *,
1640 struct dentry *, struct inode **);
1641 int vfs_rmdir(struct mnt_idmap *, struct inode *, struct dentry *);
1642 int vfs_unlink(struct mnt_idmap *, struct inode *, struct dentry *,
1646 * struct renamedata - contains all information required for renaming
1647 * @old_mnt_idmap: idmap of the old mount the inode was found from
1648 * @old_dir: parent of source
1649 * @old_dentry: source
1650 * @new_mnt_idmap: idmap of the new mount the inode was found from
1651 * @new_dir: parent of destination
1652 * @new_dentry: destination
1653 * @delegated_inode: returns an inode needing a delegation break
1654 * @flags: rename flags
1657 struct mnt_idmap *old_mnt_idmap;
1658 struct inode *old_dir;
1659 struct dentry *old_dentry;
1660 struct mnt_idmap *new_mnt_idmap;
1661 struct inode *new_dir;
1662 struct dentry *new_dentry;
1663 struct inode **delegated_inode;
1665 } __randomize_layout;
1667 int vfs_rename(struct renamedata *);
1669 static inline int vfs_whiteout(struct mnt_idmap *idmap,
1670 struct inode *dir, struct dentry *dentry)
1672 return vfs_mknod(idmap, dir, dentry, S_IFCHR | WHITEOUT_MODE,
1676 struct file *kernel_tmpfile_open(struct mnt_idmap *idmap,
1677 const struct path *parentpath,
1678 umode_t mode, int open_flag,
1679 const struct cred *cred);
1680 struct file *kernel_file_open(const struct path *path, int flags,
1681 struct inode *inode, const struct cred *cred);
1683 int vfs_mkobj(struct dentry *, umode_t,
1684 int (*f)(struct dentry *, umode_t, void *),
1687 int vfs_fchown(struct file *file, uid_t user, gid_t group);
1688 int vfs_fchmod(struct file *file, umode_t mode);
1689 int vfs_utimes(const struct path *path, struct timespec64 *times);
1691 extern long vfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1693 #ifdef CONFIG_COMPAT
1694 extern long compat_ptr_ioctl(struct file *file, unsigned int cmd,
1697 #define compat_ptr_ioctl NULL
1701 * VFS file helper functions.
1703 void inode_init_owner(struct mnt_idmap *idmap, struct inode *inode,
1704 const struct inode *dir, umode_t mode);
1705 extern bool may_open_dev(const struct path *path);
1706 umode_t mode_strip_sgid(struct mnt_idmap *idmap,
1707 const struct inode *dir, umode_t mode);
1710 * This is the "filldir" function type, used by readdir() to let
1711 * the kernel specify what kind of dirent layout it wants to have.
1712 * This allows the kernel to read directories into kernel space or
1713 * to have different dirent layouts depending on the binary type.
1714 * Return 'true' to keep going and 'false' if there are no more entries.
1717 typedef bool (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64,
1720 struct dir_context {
1726 * These flags let !MMU mmap() govern direct device mapping vs immediate
1727 * copying more easily for MAP_PRIVATE, especially for ROM filesystems.
1729 * NOMMU_MAP_COPY: Copy can be mapped (MAP_PRIVATE)
1730 * NOMMU_MAP_DIRECT: Can be mapped directly (MAP_SHARED)
1731 * NOMMU_MAP_READ: Can be mapped for reading
1732 * NOMMU_MAP_WRITE: Can be mapped for writing
1733 * NOMMU_MAP_EXEC: Can be mapped for execution
1735 #define NOMMU_MAP_COPY 0x00000001
1736 #define NOMMU_MAP_DIRECT 0x00000008
1737 #define NOMMU_MAP_READ VM_MAYREAD
1738 #define NOMMU_MAP_WRITE VM_MAYWRITE
1739 #define NOMMU_MAP_EXEC VM_MAYEXEC
1741 #define NOMMU_VMFLAGS \
1742 (NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC)
1745 * These flags control the behavior of the remap_file_range function pointer.
1746 * If it is called with len == 0 that means "remap to end of source file".
1747 * See Documentation/filesystems/vfs.rst for more details about this call.
1749 * REMAP_FILE_DEDUP: only remap if contents identical (i.e. deduplicate)
1750 * REMAP_FILE_CAN_SHORTEN: caller can handle a shortened request
1752 #define REMAP_FILE_DEDUP (1 << 0)
1753 #define REMAP_FILE_CAN_SHORTEN (1 << 1)
1756 * These flags signal that the caller is ok with altering various aspects of
1757 * the behavior of the remap operation. The changes must be made by the
1758 * implementation; the vfs remap helper functions can take advantage of them.
1759 * Flags in this category exist to preserve the quirky behavior of the hoisted
1760 * btrfs clone/dedupe ioctls.
1762 #define REMAP_FILE_ADVISORY (REMAP_FILE_CAN_SHORTEN)
1765 * These flags control the behavior of vfs_copy_file_range().
1766 * They are not available to the user via syscall.
1768 * COPY_FILE_SPLICE: call splice direct instead of fs clone/copy ops
1770 #define COPY_FILE_SPLICE (1 << 0)
1773 struct io_uring_cmd;
1775 struct file_operations {
1776 struct module *owner;
1777 loff_t (*llseek) (struct file *, loff_t, int);
1778 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
1779 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
1780 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
1781 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
1782 int (*iopoll)(struct kiocb *kiocb, struct io_comp_batch *,
1783 unsigned int flags);
1784 int (*iterate) (struct file *, struct dir_context *);
1785 int (*iterate_shared) (struct file *, struct dir_context *);
1786 __poll_t (*poll) (struct file *, struct poll_table_struct *);
1787 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
1788 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
1789 int (*mmap) (struct file *, struct vm_area_struct *);
1790 unsigned long mmap_supported_flags;
1791 int (*open) (struct inode *, struct file *);
1792 int (*flush) (struct file *, fl_owner_t id);
1793 int (*release) (struct inode *, struct file *);
1794 int (*fsync) (struct file *, loff_t, loff_t, int datasync);
1795 int (*fasync) (int, struct file *, int);
1796 int (*lock) (struct file *, int, struct file_lock *);
1797 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1798 int (*check_flags)(int);
1799 int (*flock) (struct file *, int, struct file_lock *);
1800 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
1801 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
1802 void (*splice_eof)(struct file *file);
1803 int (*setlease)(struct file *, long, struct file_lock **, void **);
1804 long (*fallocate)(struct file *file, int mode, loff_t offset,
1806 void (*show_fdinfo)(struct seq_file *m, struct file *f);
1808 unsigned (*mmap_capabilities)(struct file *);
1810 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *,
1811 loff_t, size_t, unsigned int);
1812 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in,
1813 struct file *file_out, loff_t pos_out,
1814 loff_t len, unsigned int remap_flags);
1815 int (*fadvise)(struct file *, loff_t, loff_t, int);
1816 int (*uring_cmd)(struct io_uring_cmd *ioucmd, unsigned int issue_flags);
1817 int (*uring_cmd_iopoll)(struct io_uring_cmd *, struct io_comp_batch *,
1818 unsigned int poll_flags);
1819 } __randomize_layout;
1821 struct inode_operations {
1822 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
1823 const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *);
1824 int (*permission) (struct mnt_idmap *, struct inode *, int);
1825 struct posix_acl * (*get_inode_acl)(struct inode *, int, bool);
1827 int (*readlink) (struct dentry *, char __user *,int);
1829 int (*create) (struct mnt_idmap *, struct inode *,struct dentry *,
1831 int (*link) (struct dentry *,struct inode *,struct dentry *);
1832 int (*unlink) (struct inode *,struct dentry *);
1833 int (*symlink) (struct mnt_idmap *, struct inode *,struct dentry *,
1835 int (*mkdir) (struct mnt_idmap *, struct inode *,struct dentry *,
1837 int (*rmdir) (struct inode *,struct dentry *);
1838 int (*mknod) (struct mnt_idmap *, struct inode *,struct dentry *,
1840 int (*rename) (struct mnt_idmap *, struct inode *, struct dentry *,
1841 struct inode *, struct dentry *, unsigned int);
1842 int (*setattr) (struct mnt_idmap *, struct dentry *, struct iattr *);
1843 int (*getattr) (struct mnt_idmap *, const struct path *,
1844 struct kstat *, u32, unsigned int);
1845 ssize_t (*listxattr) (struct dentry *, char *, size_t);
1846 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start,
1848 int (*update_time)(struct inode *, struct timespec64 *, int);
1849 int (*atomic_open)(struct inode *, struct dentry *,
1850 struct file *, unsigned open_flag,
1851 umode_t create_mode);
1852 int (*tmpfile) (struct mnt_idmap *, struct inode *,
1853 struct file *, umode_t);
1854 struct posix_acl *(*get_acl)(struct mnt_idmap *, struct dentry *,
1856 int (*set_acl)(struct mnt_idmap *, struct dentry *,
1857 struct posix_acl *, int);
1858 int (*fileattr_set)(struct mnt_idmap *idmap,
1859 struct dentry *dentry, struct fileattr *fa);
1860 int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa);
1861 } ____cacheline_aligned;
1863 static inline ssize_t call_read_iter(struct file *file, struct kiocb *kio,
1864 struct iov_iter *iter)
1866 return file->f_op->read_iter(kio, iter);
1869 static inline ssize_t call_write_iter(struct file *file, struct kiocb *kio,
1870 struct iov_iter *iter)
1872 return file->f_op->write_iter(kio, iter);
1875 static inline int call_mmap(struct file *file, struct vm_area_struct *vma)
1877 return file->f_op->mmap(file, vma);
1880 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
1881 extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *);
1882 extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *,
1883 loff_t, size_t, unsigned int);
1884 extern ssize_t generic_copy_file_range(struct file *file_in, loff_t pos_in,
1885 struct file *file_out, loff_t pos_out,
1886 size_t len, unsigned int flags);
1887 int __generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
1888 struct file *file_out, loff_t pos_out,
1889 loff_t *len, unsigned int remap_flags,
1890 const struct iomap_ops *dax_read_ops);
1891 int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
1892 struct file *file_out, loff_t pos_out,
1893 loff_t *count, unsigned int remap_flags);
1894 extern loff_t do_clone_file_range(struct file *file_in, loff_t pos_in,
1895 struct file *file_out, loff_t pos_out,
1896 loff_t len, unsigned int remap_flags);
1897 extern loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in,
1898 struct file *file_out, loff_t pos_out,
1899 loff_t len, unsigned int remap_flags);
1900 extern int vfs_dedupe_file_range(struct file *file,
1901 struct file_dedupe_range *same);
1902 extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos,
1903 struct file *dst_file, loff_t dst_pos,
1904 loff_t len, unsigned int remap_flags);
1907 struct super_operations {
1908 struct inode *(*alloc_inode)(struct super_block *sb);
1909 void (*destroy_inode)(struct inode *);
1910 void (*free_inode)(struct inode *);
1912 void (*dirty_inode) (struct inode *, int flags);
1913 int (*write_inode) (struct inode *, struct writeback_control *wbc);
1914 int (*drop_inode) (struct inode *);
1915 void (*evict_inode) (struct inode *);
1916 void (*put_super) (struct super_block *);
1917 int (*sync_fs)(struct super_block *sb, int wait);
1918 int (*freeze_super) (struct super_block *);
1919 int (*freeze_fs) (struct super_block *);
1920 int (*thaw_super) (struct super_block *);
1921 int (*unfreeze_fs) (struct super_block *);
1922 int (*statfs) (struct dentry *, struct kstatfs *);
1923 int (*remount_fs) (struct super_block *, int *, char *);
1924 void (*umount_begin) (struct super_block *);
1926 int (*show_options)(struct seq_file *, struct dentry *);
1927 int (*show_devname)(struct seq_file *, struct dentry *);
1928 int (*show_path)(struct seq_file *, struct dentry *);
1929 int (*show_stats)(struct seq_file *, struct dentry *);
1931 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
1932 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
1933 struct dquot **(*get_dquots)(struct inode *);
1935 long (*nr_cached_objects)(struct super_block *,
1936 struct shrink_control *);
1937 long (*free_cached_objects)(struct super_block *,
1938 struct shrink_control *);
1939 void (*shutdown)(struct super_block *sb);
1943 * Inode flags - they have no relation to superblock flags now
1945 #define S_SYNC (1 << 0) /* Writes are synced at once */
1946 #define S_NOATIME (1 << 1) /* Do not update access times */
1947 #define S_APPEND (1 << 2) /* Append-only file */
1948 #define S_IMMUTABLE (1 << 3) /* Immutable file */
1949 #define S_DEAD (1 << 4) /* removed, but still open directory */
1950 #define S_NOQUOTA (1 << 5) /* Inode is not counted to quota */
1951 #define S_DIRSYNC (1 << 6) /* Directory modifications are synchronous */
1952 #define S_NOCMTIME (1 << 7) /* Do not update file c/mtime */
1953 #define S_SWAPFILE (1 << 8) /* Do not truncate: swapon got its bmaps */
1954 #define S_PRIVATE (1 << 9) /* Inode is fs-internal */
1955 #define S_IMA (1 << 10) /* Inode has an associated IMA struct */
1956 #define S_AUTOMOUNT (1 << 11) /* Automount/referral quasi-directory */
1957 #define S_NOSEC (1 << 12) /* no suid or xattr security attributes */
1958 #ifdef CONFIG_FS_DAX
1959 #define S_DAX (1 << 13) /* Direct Access, avoiding the page cache */
1961 #define S_DAX 0 /* Make all the DAX code disappear */
1963 #define S_ENCRYPTED (1 << 14) /* Encrypted file (using fs/crypto/) */
1964 #define S_CASEFOLD (1 << 15) /* Casefolded file */
1965 #define S_VERITY (1 << 16) /* Verity file (using fs/verity/) */
1966 #define S_KERNEL_FILE (1 << 17) /* File is in use by the kernel (eg. fs/cachefiles) */
1969 * Note that nosuid etc flags are inode-specific: setting some file-system
1970 * flags just means all the inodes inherit those flags by default. It might be
1971 * possible to override it selectively if you really wanted to with some
1972 * ioctl() that is not currently implemented.
1974 * Exception: SB_RDONLY is always applied to the entire file system.
1976 * Unfortunately, it is possible to change a filesystems flags with it mounted
1977 * with files in use. This means that all of the inodes will not have their
1978 * i_flags updated. Hence, i_flags no longer inherit the superblock mount
1979 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org
1981 #define __IS_FLG(inode, flg) ((inode)->i_sb->s_flags & (flg))
1983 static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; }
1984 #define IS_RDONLY(inode) sb_rdonly((inode)->i_sb)
1985 #define IS_SYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS) || \
1986 ((inode)->i_flags & S_SYNC))
1987 #define IS_DIRSYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS|SB_DIRSYNC) || \
1988 ((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
1989 #define IS_MANDLOCK(inode) __IS_FLG(inode, SB_MANDLOCK)
1990 #define IS_NOATIME(inode) __IS_FLG(inode, SB_RDONLY|SB_NOATIME)
1991 #define IS_I_VERSION(inode) __IS_FLG(inode, SB_I_VERSION)
1993 #define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA)
1994 #define IS_APPEND(inode) ((inode)->i_flags & S_APPEND)
1995 #define IS_IMMUTABLE(inode) ((inode)->i_flags & S_IMMUTABLE)
1996 #define IS_POSIXACL(inode) __IS_FLG(inode, SB_POSIXACL)
1998 #define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD)
1999 #define IS_NOCMTIME(inode) ((inode)->i_flags & S_NOCMTIME)
2000 #define IS_SWAPFILE(inode) ((inode)->i_flags & S_SWAPFILE)
2001 #define IS_PRIVATE(inode) ((inode)->i_flags & S_PRIVATE)
2002 #define IS_IMA(inode) ((inode)->i_flags & S_IMA)
2003 #define IS_AUTOMOUNT(inode) ((inode)->i_flags & S_AUTOMOUNT)
2004 #define IS_NOSEC(inode) ((inode)->i_flags & S_NOSEC)
2005 #define IS_DAX(inode) ((inode)->i_flags & S_DAX)
2006 #define IS_ENCRYPTED(inode) ((inode)->i_flags & S_ENCRYPTED)
2007 #define IS_CASEFOLDED(inode) ((inode)->i_flags & S_CASEFOLD)
2008 #define IS_VERITY(inode) ((inode)->i_flags & S_VERITY)
2010 #define IS_WHITEOUT(inode) (S_ISCHR(inode->i_mode) && \
2011 (inode)->i_rdev == WHITEOUT_DEV)
2013 static inline bool HAS_UNMAPPED_ID(struct mnt_idmap *idmap,
2014 struct inode *inode)
2016 return !vfsuid_valid(i_uid_into_vfsuid(idmap, inode)) ||
2017 !vfsgid_valid(i_gid_into_vfsgid(idmap, inode));
2020 static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp)
2022 *kiocb = (struct kiocb) {
2024 .ki_flags = filp->f_iocb_flags,
2025 .ki_ioprio = get_current_ioprio(),
2029 static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
2032 *kiocb = (struct kiocb) {
2034 .ki_flags = kiocb_src->ki_flags,
2035 .ki_ioprio = kiocb_src->ki_ioprio,
2036 .ki_pos = kiocb_src->ki_pos,
2041 * Inode state bits. Protected by inode->i_lock
2043 * Four bits determine the dirty state of the inode: I_DIRTY_SYNC,
2044 * I_DIRTY_DATASYNC, I_DIRTY_PAGES, and I_DIRTY_TIME.
2046 * Four bits define the lifetime of an inode. Initially, inodes are I_NEW,
2047 * until that flag is cleared. I_WILL_FREE, I_FREEING and I_CLEAR are set at
2048 * various stages of removing an inode.
2050 * Two bits are used for locking and completion notification, I_NEW and I_SYNC.
2052 * I_DIRTY_SYNC Inode is dirty, but doesn't have to be written on
2053 * fdatasync() (unless I_DIRTY_DATASYNC is also set).
2054 * Timestamp updates are the usual cause.
2055 * I_DIRTY_DATASYNC Data-related inode changes pending. We keep track of
2056 * these changes separately from I_DIRTY_SYNC so that we
2057 * don't have to write inode on fdatasync() when only
2058 * e.g. the timestamps have changed.
2059 * I_DIRTY_PAGES Inode has dirty pages. Inode itself may be clean.
2060 * I_DIRTY_TIME The inode itself has dirty timestamps, and the
2061 * lazytime mount option is enabled. We keep track of this
2062 * separately from I_DIRTY_SYNC in order to implement
2063 * lazytime. This gets cleared if I_DIRTY_INODE
2064 * (I_DIRTY_SYNC and/or I_DIRTY_DATASYNC) gets set. But
2065 * I_DIRTY_TIME can still be set if I_DIRTY_SYNC is already
2066 * in place because writeback might already be in progress
2067 * and we don't want to lose the time update
2068 * I_NEW Serves as both a mutex and completion notification.
2069 * New inodes set I_NEW. If two processes both create
2070 * the same inode, one of them will release its inode and
2071 * wait for I_NEW to be released before returning.
2072 * Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can
2073 * also cause waiting on I_NEW, without I_NEW actually
2074 * being set. find_inode() uses this to prevent returning
2075 * nearly-dead inodes.
2076 * I_WILL_FREE Must be set when calling write_inode_now() if i_count
2077 * is zero. I_FREEING must be set when I_WILL_FREE is
2079 * I_FREEING Set when inode is about to be freed but still has dirty
2080 * pages or buffers attached or the inode itself is still
2082 * I_CLEAR Added by clear_inode(). In this state the inode is
2083 * clean and can be destroyed. Inode keeps I_FREEING.
2085 * Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are
2086 * prohibited for many purposes. iget() must wait for
2087 * the inode to be completely released, then create it
2088 * anew. Other functions will just ignore such inodes,
2089 * if appropriate. I_NEW is used for waiting.
2091 * I_SYNC Writeback of inode is running. The bit is set during
2092 * data writeback, and cleared with a wakeup on the bit
2093 * address once it is done. The bit is also used to pin
2094 * the inode in memory for flusher thread.
2096 * I_REFERENCED Marks the inode as recently references on the LRU list.
2098 * I_DIO_WAKEUP Never set. Only used as a key for wait_on_bit().
2100 * I_WB_SWITCH Cgroup bdi_writeback switching in progress. Used to
2101 * synchronize competing switching instances and to tell
2102 * wb stat updates to grab the i_pages lock. See
2103 * inode_switch_wbs_work_fn() for details.
2105 * I_OVL_INUSE Used by overlayfs to get exclusive ownership on upper
2106 * and work dirs among overlayfs mounts.
2108 * I_CREATING New object's inode in the middle of setting up.
2110 * I_DONTCACHE Evict inode as soon as it is not used anymore.
2112 * I_SYNC_QUEUED Inode is queued in b_io or b_more_io writeback lists.
2113 * Used to detect that mark_inode_dirty() should not move
2114 * inode between dirty lists.
2116 * I_PINNING_FSCACHE_WB Inode is pinning an fscache object for writeback.
2118 * Q: What is the difference between I_WILL_FREE and I_FREEING?
2120 #define I_DIRTY_SYNC (1 << 0)
2121 #define I_DIRTY_DATASYNC (1 << 1)
2122 #define I_DIRTY_PAGES (1 << 2)
2124 #define I_NEW (1 << __I_NEW)
2125 #define I_WILL_FREE (1 << 4)
2126 #define I_FREEING (1 << 5)
2127 #define I_CLEAR (1 << 6)
2129 #define I_SYNC (1 << __I_SYNC)
2130 #define I_REFERENCED (1 << 8)
2131 #define __I_DIO_WAKEUP 9
2132 #define I_DIO_WAKEUP (1 << __I_DIO_WAKEUP)
2133 #define I_LINKABLE (1 << 10)
2134 #define I_DIRTY_TIME (1 << 11)
2135 #define I_WB_SWITCH (1 << 13)
2136 #define I_OVL_INUSE (1 << 14)
2137 #define I_CREATING (1 << 15)
2138 #define I_DONTCACHE (1 << 16)
2139 #define I_SYNC_QUEUED (1 << 17)
2140 #define I_PINNING_FSCACHE_WB (1 << 18)
2142 #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC)
2143 #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES)
2144 #define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME)
2146 extern void __mark_inode_dirty(struct inode *, int);
2147 static inline void mark_inode_dirty(struct inode *inode)
2149 __mark_inode_dirty(inode, I_DIRTY);
2152 static inline void mark_inode_dirty_sync(struct inode *inode)
2154 __mark_inode_dirty(inode, I_DIRTY_SYNC);
2158 * Returns true if the given inode itself only has dirty timestamps (its pages
2159 * may still be dirty) and isn't currently being allocated or freed.
2160 * Filesystems should call this if when writing an inode when lazytime is
2161 * enabled, they want to opportunistically write the timestamps of other inodes
2162 * located very nearby on-disk, e.g. in the same inode block. This returns true
2163 * if the given inode is in need of such an opportunistic update. Requires
2164 * i_lock, or at least later re-checking under i_lock.
2166 static inline bool inode_is_dirtytime_only(struct inode *inode)
2168 return (inode->i_state & (I_DIRTY_TIME | I_NEW |
2169 I_FREEING | I_WILL_FREE)) == I_DIRTY_TIME;
2172 extern void inc_nlink(struct inode *inode);
2173 extern void drop_nlink(struct inode *inode);
2174 extern void clear_nlink(struct inode *inode);
2175 extern void set_nlink(struct inode *inode, unsigned int nlink);
2177 static inline void inode_inc_link_count(struct inode *inode)
2180 mark_inode_dirty(inode);
2183 static inline void inode_dec_link_count(struct inode *inode)
2186 mark_inode_dirty(inode);
2189 enum file_time_flags {
2196 extern bool atime_needs_update(const struct path *, struct inode *);
2197 extern void touch_atime(const struct path *);
2198 int inode_update_time(struct inode *inode, struct timespec64 *time, int flags);
2200 static inline void file_accessed(struct file *file)
2202 if (!(file->f_flags & O_NOATIME))
2203 touch_atime(&file->f_path);
2206 extern int file_modified(struct file *file);
2207 int kiocb_modified(struct kiocb *iocb);
2209 int sync_inode_metadata(struct inode *inode, int wait);
2211 struct file_system_type {
2214 #define FS_REQUIRES_DEV 1
2215 #define FS_BINARY_MOUNTDATA 2
2216 #define FS_HAS_SUBTYPE 4
2217 #define FS_USERNS_MOUNT 8 /* Can be mounted by userns root */
2218 #define FS_DISALLOW_NOTIFY_PERM 16 /* Disable fanotify permission events */
2219 #define FS_ALLOW_IDMAP 32 /* FS has been updated to handle vfs idmappings. */
2220 #define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() during rename() internally. */
2221 int (*init_fs_context)(struct fs_context *);
2222 const struct fs_parameter_spec *parameters;
2223 struct dentry *(*mount) (struct file_system_type *, int,
2224 const char *, void *);
2225 void (*kill_sb) (struct super_block *);
2226 struct module *owner;
2227 struct file_system_type * next;
2228 struct hlist_head fs_supers;
2230 struct lock_class_key s_lock_key;
2231 struct lock_class_key s_umount_key;
2232 struct lock_class_key s_vfs_rename_key;
2233 struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
2235 struct lock_class_key i_lock_key;
2236 struct lock_class_key i_mutex_key;
2237 struct lock_class_key invalidate_lock_key;
2238 struct lock_class_key i_mutex_dir_key;
2241 #define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME)
2243 extern struct dentry *mount_bdev(struct file_system_type *fs_type,
2244 int flags, const char *dev_name, void *data,
2245 int (*fill_super)(struct super_block *, void *, int));
2246 extern struct dentry *mount_single(struct file_system_type *fs_type,
2247 int flags, void *data,
2248 int (*fill_super)(struct super_block *, void *, int));
2249 extern struct dentry *mount_nodev(struct file_system_type *fs_type,
2250 int flags, void *data,
2251 int (*fill_super)(struct super_block *, void *, int));
2252 extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path);
2253 void retire_super(struct super_block *sb);
2254 void generic_shutdown_super(struct super_block *sb);
2255 void kill_block_super(struct super_block *sb);
2256 void kill_anon_super(struct super_block *sb);
2257 void kill_litter_super(struct super_block *sb);
2258 void deactivate_super(struct super_block *sb);
2259 void deactivate_locked_super(struct super_block *sb);
2260 int set_anon_super(struct super_block *s, void *data);
2261 int set_anon_super_fc(struct super_block *s, struct fs_context *fc);
2262 int get_anon_bdev(dev_t *);
2263 void free_anon_bdev(dev_t);
2264 struct super_block *sget_fc(struct fs_context *fc,
2265 int (*test)(struct super_block *, struct fs_context *),
2266 int (*set)(struct super_block *, struct fs_context *));
2267 struct super_block *sget(struct file_system_type *type,
2268 int (*test)(struct super_block *,void *),
2269 int (*set)(struct super_block *,void *),
2270 int flags, void *data);
2272 /* Alas, no aliases. Too much hassle with bringing module.h everywhere */
2273 #define fops_get(fops) \
2274 (((fops) && try_module_get((fops)->owner) ? (fops) : NULL))
2275 #define fops_put(fops) \
2276 do { if (fops) module_put((fops)->owner); } while(0)
2278 * This one is to be used *ONLY* from ->open() instances.
2279 * fops must be non-NULL, pinned down *and* module dependencies
2280 * should be sufficient to pin the caller down as well.
2282 #define replace_fops(f, fops) \
2284 struct file *__file = (f); \
2285 fops_put(__file->f_op); \
2286 BUG_ON(!(__file->f_op = (fops))); \
2289 extern int register_filesystem(struct file_system_type *);
2290 extern int unregister_filesystem(struct file_system_type *);
2291 extern int vfs_statfs(const struct path *, struct kstatfs *);
2292 extern int user_statfs(const char __user *, struct kstatfs *);
2293 extern int fd_statfs(int, struct kstatfs *);
2294 extern int freeze_super(struct super_block *super);
2295 extern int thaw_super(struct super_block *super);
2296 extern __printf(2, 3)
2297 int super_setup_bdi_name(struct super_block *sb, char *fmt, ...);
2298 extern int super_setup_bdi(struct super_block *sb);
2300 extern int current_umask(void);
2302 extern void ihold(struct inode * inode);
2303 extern void iput(struct inode *);
2304 extern int generic_update_time(struct inode *, struct timespec64 *, int);
2307 extern struct kobject *fs_kobj;
2309 #define MAX_RW_COUNT (INT_MAX & PAGE_MASK)
2314 const char *name; /* pointer to actual string */
2315 const __user char *uptr; /* original userland pointer */
2317 struct audit_names *aname;
2320 static_assert(offsetof(struct filename, iname) % sizeof(long) == 0);
2322 static inline struct mnt_idmap *file_mnt_idmap(struct file *file)
2324 return mnt_idmap(file->f_path.mnt);
2328 * is_idmapped_mnt - check whether a mount is mapped
2329 * @mnt: the mount to check
2331 * If @mnt has an non @nop_mnt_idmap attached to it then @mnt is mapped.
2333 * Return: true if mount is mapped, false if not.
2335 static inline bool is_idmapped_mnt(const struct vfsmount *mnt)
2337 return mnt_idmap(mnt) != &nop_mnt_idmap;
2340 extern long vfs_truncate(const struct path *, loff_t);
2341 int do_truncate(struct mnt_idmap *, struct dentry *, loff_t start,
2342 unsigned int time_attrs, struct file *filp);
2343 extern int vfs_fallocate(struct file *file, int mode, loff_t offset,
2345 extern long do_sys_open(int dfd, const char __user *filename, int flags,
2347 extern struct file *file_open_name(struct filename *, int, umode_t);
2348 extern struct file *filp_open(const char *, int, umode_t);
2349 extern struct file *file_open_root(const struct path *,
2350 const char *, int, umode_t);
2351 static inline struct file *file_open_root_mnt(struct vfsmount *mnt,
2352 const char *name, int flags, umode_t mode)
2354 return file_open_root(&(struct path){.mnt = mnt, .dentry = mnt->mnt_root},
2357 struct file *dentry_open(const struct path *path, int flags,
2358 const struct cred *creds);
2359 struct file *dentry_create(const struct path *path, int flags, umode_t mode,
2360 const struct cred *cred);
2361 struct file *backing_file_open(const struct path *path, int flags,
2362 const struct path *real_path,
2363 const struct cred *cred);
2364 struct path *backing_file_real_path(struct file *f);
2367 * file_real_path - get the path corresponding to f_inode
2369 * When opening a backing file for a stackable filesystem (e.g.,
2370 * overlayfs) f_path may be on the stackable filesystem and f_inode on
2371 * the underlying filesystem. When the path associated with f_inode is
2372 * needed, this helper should be used instead of accessing f_path
2375 static inline const struct path *file_real_path(struct file *f)
2377 if (unlikely(f->f_mode & FMODE_BACKING))
2378 return backing_file_real_path(f);
2382 static inline struct file *file_clone_open(struct file *file)
2384 return dentry_open(&file->f_path, file->f_flags, file->f_cred);
2386 extern int filp_close(struct file *, fl_owner_t id);
2388 extern struct filename *getname_flags(const char __user *, int, int *);
2389 extern struct filename *getname_uflags(const char __user *, int);
2390 extern struct filename *getname(const char __user *);
2391 extern struct filename *getname_kernel(const char *);
2392 extern void putname(struct filename *name);
2394 extern int finish_open(struct file *file, struct dentry *dentry,
2395 int (*open)(struct inode *, struct file *));
2396 extern int finish_no_open(struct file *file, struct dentry *dentry);
2398 /* Helper for the simple case when original dentry is used */
2399 static inline int finish_open_simple(struct file *file, int error)
2404 return finish_open(file, file->f_path.dentry, NULL);
2408 extern void __init vfs_caches_init_early(void);
2409 extern void __init vfs_caches_init(void);
2411 extern struct kmem_cache *names_cachep;
2413 #define __getname() kmem_cache_alloc(names_cachep, GFP_KERNEL)
2414 #define __putname(name) kmem_cache_free(names_cachep, (void *)(name))
2416 extern struct super_block *blockdev_superblock;
2417 static inline bool sb_is_blkdev_sb(struct super_block *sb)
2419 return IS_ENABLED(CONFIG_BLOCK) && sb == blockdev_superblock;
2422 void emergency_thaw_all(void);
2423 extern int sync_filesystem(struct super_block *);
2424 extern const struct file_operations def_blk_fops;
2425 extern const struct file_operations def_chr_fops;
2428 #define CHRDEV_MAJOR_MAX 512
2429 /* Marks the bottom of the first segment of free char majors */
2430 #define CHRDEV_MAJOR_DYN_END 234
2431 /* Marks the top and bottom of the second segment of free char majors */
2432 #define CHRDEV_MAJOR_DYN_EXT_START 511
2433 #define CHRDEV_MAJOR_DYN_EXT_END 384
2435 extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *);
2436 extern int register_chrdev_region(dev_t, unsigned, const char *);
2437 extern int __register_chrdev(unsigned int major, unsigned int baseminor,
2438 unsigned int count, const char *name,
2439 const struct file_operations *fops);
2440 extern void __unregister_chrdev(unsigned int major, unsigned int baseminor,
2441 unsigned int count, const char *name);
2442 extern void unregister_chrdev_region(dev_t, unsigned);
2443 extern void chrdev_show(struct seq_file *,off_t);
2445 static inline int register_chrdev(unsigned int major, const char *name,
2446 const struct file_operations *fops)
2448 return __register_chrdev(major, 0, 256, name, fops);
2451 static inline void unregister_chrdev(unsigned int major, const char *name)
2453 __unregister_chrdev(major, 0, 256, name);
2456 extern void init_special_inode(struct inode *, umode_t, dev_t);
2458 /* Invalid inode operations -- fs/bad_inode.c */
2459 extern void make_bad_inode(struct inode *);
2460 extern bool is_bad_inode(struct inode *);
2462 extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart,
2464 extern int __must_check file_check_and_advance_wb_err(struct file *file);
2465 extern int __must_check file_write_and_wait_range(struct file *file,
2466 loff_t start, loff_t end);
2468 static inline int file_write_and_wait(struct file *file)
2470 return file_write_and_wait_range(file, 0, LLONG_MAX);
2473 extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end,
2475 extern int vfs_fsync(struct file *file, int datasync);
2477 extern int sync_file_range(struct file *file, loff_t offset, loff_t nbytes,
2478 unsigned int flags);
2480 static inline bool iocb_is_dsync(const struct kiocb *iocb)
2482 return (iocb->ki_flags & IOCB_DSYNC) ||
2483 IS_SYNC(iocb->ki_filp->f_mapping->host);
2487 * Sync the bytes written if this was a synchronous write. Expect ki_pos
2488 * to already be updated for the write, and will return either the amount
2489 * of bytes passed in, or an error if syncing the file failed.
2491 static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count)
2493 if (iocb_is_dsync(iocb)) {
2494 int ret = vfs_fsync_range(iocb->ki_filp,
2495 iocb->ki_pos - count, iocb->ki_pos - 1,
2496 (iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
2504 extern void emergency_sync(void);
2505 extern void emergency_remount(void);
2508 extern int bmap(struct inode *inode, sector_t *block);
2510 static inline int bmap(struct inode *inode, sector_t *block)
2516 int notify_change(struct mnt_idmap *, struct dentry *,
2517 struct iattr *, struct inode **);
2518 int inode_permission(struct mnt_idmap *, struct inode *, int);
2519 int generic_permission(struct mnt_idmap *, struct inode *, int);
2520 static inline int file_permission(struct file *file, int mask)
2522 return inode_permission(file_mnt_idmap(file),
2523 file_inode(file), mask);
2525 static inline int path_permission(const struct path *path, int mask)
2527 return inode_permission(mnt_idmap(path->mnt),
2528 d_inode(path->dentry), mask);
2530 int __check_sticky(struct mnt_idmap *idmap, struct inode *dir,
2531 struct inode *inode);
2533 static inline bool execute_ok(struct inode *inode)
2535 return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode);
2538 static inline bool inode_wrong_type(const struct inode *inode, umode_t mode)
2540 return (inode->i_mode ^ mode) & S_IFMT;
2543 static inline void file_start_write(struct file *file)
2545 if (!S_ISREG(file_inode(file)->i_mode))
2547 sb_start_write(file_inode(file)->i_sb);
2550 static inline bool file_start_write_trylock(struct file *file)
2552 if (!S_ISREG(file_inode(file)->i_mode))
2554 return sb_start_write_trylock(file_inode(file)->i_sb);
2557 static inline void file_end_write(struct file *file)
2559 if (!S_ISREG(file_inode(file)->i_mode))
2561 __sb_end_write(file_inode(file)->i_sb, SB_FREEZE_WRITE);
2565 * This is used for regular files where some users -- especially the
2566 * currently executed binary in a process, previously handled via
2567 * VM_DENYWRITE -- cannot handle concurrent write (and maybe mmap
2568 * read-write shared) accesses.
2570 * get_write_access() gets write permission for a file.
2571 * put_write_access() releases this write permission.
2572 * deny_write_access() denies write access to a file.
2573 * allow_write_access() re-enables write access to a file.
2575 * The i_writecount field of an inode can have the following values:
2576 * 0: no write access, no denied write access
2577 * < 0: (-i_writecount) users that denied write access to the file.
2578 * > 0: (i_writecount) users that have write access to the file.
2580 * Normally we operate on that counter with atomic_{inc,dec} and it's safe
2581 * except for the cases where we don't hold i_writecount yet. Then we need to
2582 * use {get,deny}_write_access() - these functions check the sign and refuse
2583 * to do the change if sign is wrong.
2585 static inline int get_write_access(struct inode *inode)
2587 return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY;
2589 static inline int deny_write_access(struct file *file)
2591 struct inode *inode = file_inode(file);
2592 return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY;
2594 static inline void put_write_access(struct inode * inode)
2596 atomic_dec(&inode->i_writecount);
2598 static inline void allow_write_access(struct file *file)
2601 atomic_inc(&file_inode(file)->i_writecount);
2603 static inline bool inode_is_open_for_write(const struct inode *inode)
2605 return atomic_read(&inode->i_writecount) > 0;
2608 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
2609 static inline void i_readcount_dec(struct inode *inode)
2611 BUG_ON(!atomic_read(&inode->i_readcount));
2612 atomic_dec(&inode->i_readcount);
2614 static inline void i_readcount_inc(struct inode *inode)
2616 atomic_inc(&inode->i_readcount);
2619 static inline void i_readcount_dec(struct inode *inode)
2623 static inline void i_readcount_inc(struct inode *inode)
2628 extern int do_pipe_flags(int *, int);
2630 extern ssize_t kernel_read(struct file *, void *, size_t, loff_t *);
2631 ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos);
2632 extern ssize_t kernel_write(struct file *, const void *, size_t, loff_t *);
2633 extern ssize_t __kernel_write(struct file *, const void *, size_t, loff_t *);
2634 extern struct file * open_exec(const char *);
2636 /* fs/dcache.c -- generic fs support functions */
2637 extern bool is_subdir(struct dentry *, struct dentry *);
2638 extern bool path_is_under(const struct path *, const struct path *);
2640 extern char *file_path(struct file *, char *, int);
2642 #include <linux/err.h>
2644 /* needed for stackable file system support */
2645 extern loff_t default_llseek(struct file *file, loff_t offset, int whence);
2647 extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence);
2649 extern int inode_init_always(struct super_block *, struct inode *);
2650 extern void inode_init_once(struct inode *);
2651 extern void address_space_init_once(struct address_space *mapping);
2652 extern struct inode * igrab(struct inode *);
2653 extern ino_t iunique(struct super_block *, ino_t);
2654 extern int inode_needs_sync(struct inode *inode);
2655 extern int generic_delete_inode(struct inode *inode);
2656 static inline int generic_drop_inode(struct inode *inode)
2658 return !inode->i_nlink || inode_unhashed(inode);
2660 extern void d_mark_dontcache(struct inode *inode);
2662 extern struct inode *ilookup5_nowait(struct super_block *sb,
2663 unsigned long hashval, int (*test)(struct inode *, void *),
2665 extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
2666 int (*test)(struct inode *, void *), void *data);
2667 extern struct inode *ilookup(struct super_block *sb, unsigned long ino);
2669 extern struct inode *inode_insert5(struct inode *inode, unsigned long hashval,
2670 int (*test)(struct inode *, void *),
2671 int (*set)(struct inode *, void *),
2673 extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *);
2674 extern struct inode * iget_locked(struct super_block *, unsigned long);
2675 extern struct inode *find_inode_nowait(struct super_block *,
2677 int (*match)(struct inode *,
2678 unsigned long, void *),
2680 extern struct inode *find_inode_rcu(struct super_block *, unsigned long,
2681 int (*)(struct inode *, void *), void *);
2682 extern struct inode *find_inode_by_ino_rcu(struct super_block *, unsigned long);
2683 extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *);
2684 extern int insert_inode_locked(struct inode *);
2685 #ifdef CONFIG_DEBUG_LOCK_ALLOC
2686 extern void lockdep_annotate_inode_mutex_key(struct inode *inode);
2688 static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { };
2690 extern void unlock_new_inode(struct inode *);
2691 extern void discard_new_inode(struct inode *);
2692 extern unsigned int get_next_ino(void);
2693 extern void evict_inodes(struct super_block *sb);
2694 void dump_mapping(const struct address_space *);
2697 * Userspace may rely on the inode number being non-zero. For example, glibc
2698 * simply ignores files with zero i_ino in unlink() and other places.
2700 * As an additional complication, if userspace was compiled with
2701 * _FILE_OFFSET_BITS=32 on a 64-bit kernel we'll only end up reading out the
2702 * lower 32 bits, so we need to check that those aren't zero explicitly. With
2703 * _FILE_OFFSET_BITS=64, this may cause some harmless false-negatives, but
2704 * better safe than sorry.
2706 static inline bool is_zero_ino(ino_t ino)
2708 return (u32)ino == 0;
2711 extern void __iget(struct inode * inode);
2712 extern void iget_failed(struct inode *);
2713 extern void clear_inode(struct inode *);
2714 extern void __destroy_inode(struct inode *);
2715 extern struct inode *new_inode_pseudo(struct super_block *sb);
2716 extern struct inode *new_inode(struct super_block *sb);
2717 extern void free_inode_nonrcu(struct inode *inode);
2718 extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *);
2719 extern int file_remove_privs(struct file *);
2720 int setattr_should_drop_sgid(struct mnt_idmap *idmap,
2721 const struct inode *inode);
2724 * This must be used for allocating filesystems specific inodes to set
2725 * up the inode reclaim context correctly.
2727 static inline void *
2728 alloc_inode_sb(struct super_block *sb, struct kmem_cache *cache, gfp_t gfp)
2730 return kmem_cache_alloc_lru(cache, &sb->s_inode_lru, gfp);
2733 extern void __insert_inode_hash(struct inode *, unsigned long hashval);
2734 static inline void insert_inode_hash(struct inode *inode)
2736 __insert_inode_hash(inode, inode->i_ino);
2739 extern void __remove_inode_hash(struct inode *);
2740 static inline void remove_inode_hash(struct inode *inode)
2742 if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash))
2743 __remove_inode_hash(inode);
2746 extern void inode_sb_list_add(struct inode *inode);
2747 extern void inode_add_lru(struct inode *inode);
2749 extern int sb_set_blocksize(struct super_block *, int);
2750 extern int sb_min_blocksize(struct super_block *, int);
2752 extern int generic_file_mmap(struct file *, struct vm_area_struct *);
2753 extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *);
2754 extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *);
2755 int generic_write_checks_count(struct kiocb *iocb, loff_t *count);
2756 extern int generic_write_check_limits(struct file *file, loff_t pos,
2758 extern int generic_file_rw_checks(struct file *file_in, struct file *file_out);
2759 ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *to,
2760 ssize_t already_read);
2761 extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *);
2762 extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *);
2763 extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *);
2764 extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *);
2765 ssize_t generic_perform_write(struct kiocb *, struct iov_iter *);
2766 ssize_t direct_write_fallback(struct kiocb *iocb, struct iov_iter *iter,
2767 ssize_t direct_written, ssize_t buffered_written);
2769 ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos,
2771 ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos,
2773 ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb,
2774 struct iov_iter *iter);
2775 ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb,
2776 struct iov_iter *iter);
2779 ssize_t filemap_splice_read(struct file *in, loff_t *ppos,
2780 struct pipe_inode_info *pipe,
2781 size_t len, unsigned int flags);
2782 ssize_t copy_splice_read(struct file *in, loff_t *ppos,
2783 struct pipe_inode_info *pipe,
2784 size_t len, unsigned int flags);
2785 extern ssize_t iter_file_splice_write(struct pipe_inode_info *,
2786 struct file *, loff_t *, size_t, unsigned int);
2787 extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
2788 loff_t *opos, size_t len, unsigned int flags);
2792 file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping);
2793 extern loff_t noop_llseek(struct file *file, loff_t offset, int whence);
2794 #define no_llseek NULL
2795 extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize);
2796 extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence);
2797 extern loff_t generic_file_llseek_size(struct file *file, loff_t offset,
2798 int whence, loff_t maxsize, loff_t eof);
2799 extern loff_t fixed_size_llseek(struct file *file, loff_t offset,
2800 int whence, loff_t size);
2801 extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t);
2802 extern loff_t no_seek_end_llseek(struct file *, loff_t, int);
2803 int rw_verify_area(int, struct file *, const loff_t *, size_t);
2804 extern int generic_file_open(struct inode * inode, struct file * filp);
2805 extern int nonseekable_open(struct inode * inode, struct file * filp);
2806 extern int stream_open(struct inode * inode, struct file * filp);
2809 typedef void (dio_submit_t)(struct bio *bio, struct inode *inode,
2810 loff_t file_offset);
2813 /* need locking between buffered and direct access */
2816 /* filesystem does not support filling holes */
2817 DIO_SKIP_HOLES = 0x02,
2820 ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
2821 struct block_device *bdev, struct iov_iter *iter,
2822 get_block_t get_block,
2823 dio_iodone_t end_io,
2826 static inline ssize_t blockdev_direct_IO(struct kiocb *iocb,
2827 struct inode *inode,
2828 struct iov_iter *iter,
2829 get_block_t get_block)
2831 return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter,
2832 get_block, NULL, DIO_LOCKING | DIO_SKIP_HOLES);
2836 void inode_dio_wait(struct inode *inode);
2839 * inode_dio_begin - signal start of a direct I/O requests
2840 * @inode: inode the direct I/O happens on
2842 * This is called once we've finished processing a direct I/O request,
2843 * and is used to wake up callers waiting for direct I/O to be quiesced.
2845 static inline void inode_dio_begin(struct inode *inode)
2847 atomic_inc(&inode->i_dio_count);
2851 * inode_dio_end - signal finish of a direct I/O requests
2852 * @inode: inode the direct I/O happens on
2854 * This is called once we've finished processing a direct I/O request,
2855 * and is used to wake up callers waiting for direct I/O to be quiesced.
2857 static inline void inode_dio_end(struct inode *inode)
2859 if (atomic_dec_and_test(&inode->i_dio_count))
2860 wake_up_bit(&inode->i_state, __I_DIO_WAKEUP);
2863 extern void inode_set_flags(struct inode *inode, unsigned int flags,
2866 extern const struct file_operations generic_ro_fops;
2868 #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m))
2870 extern int readlink_copy(char __user *, int, const char *);
2871 extern int page_readlink(struct dentry *, char __user *, int);
2872 extern const char *page_get_link(struct dentry *, struct inode *,
2873 struct delayed_call *);
2874 extern void page_put_link(void *);
2875 extern int page_symlink(struct inode *inode, const char *symname, int len);
2876 extern const struct inode_operations page_symlink_inode_operations;
2877 extern void kfree_link(void *);
2878 void generic_fillattr(struct mnt_idmap *, struct inode *, struct kstat *);
2879 void generic_fill_statx_attr(struct inode *inode, struct kstat *stat);
2880 extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int);
2881 extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int);
2882 void __inode_add_bytes(struct inode *inode, loff_t bytes);
2883 void inode_add_bytes(struct inode *inode, loff_t bytes);
2884 void __inode_sub_bytes(struct inode *inode, loff_t bytes);
2885 void inode_sub_bytes(struct inode *inode, loff_t bytes);
2886 static inline loff_t __inode_get_bytes(struct inode *inode)
2888 return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes;
2890 loff_t inode_get_bytes(struct inode *inode);
2891 void inode_set_bytes(struct inode *inode, loff_t bytes);
2892 const char *simple_get_link(struct dentry *, struct inode *,
2893 struct delayed_call *);
2894 extern const struct inode_operations simple_symlink_inode_operations;
2896 extern int iterate_dir(struct file *, struct dir_context *);
2898 int vfs_fstatat(int dfd, const char __user *filename, struct kstat *stat,
2900 int vfs_fstat(int fd, struct kstat *stat);
2902 static inline int vfs_stat(const char __user *filename, struct kstat *stat)
2904 return vfs_fstatat(AT_FDCWD, filename, stat, 0);
2906 static inline int vfs_lstat(const char __user *name, struct kstat *stat)
2908 return vfs_fstatat(AT_FDCWD, name, stat, AT_SYMLINK_NOFOLLOW);
2911 extern const char *vfs_get_link(struct dentry *, struct delayed_call *);
2912 extern int vfs_readlink(struct dentry *, char __user *, int);
2914 extern struct file_system_type *get_filesystem(struct file_system_type *fs);
2915 extern void put_filesystem(struct file_system_type *fs);
2916 extern struct file_system_type *get_fs_type(const char *name);
2917 extern struct super_block *get_super(struct block_device *);
2918 extern struct super_block *get_active_super(struct block_device *bdev);
2919 extern void drop_super(struct super_block *sb);
2920 extern void drop_super_exclusive(struct super_block *sb);
2921 extern void iterate_supers(void (*)(struct super_block *, void *), void *);
2922 extern void iterate_supers_type(struct file_system_type *,
2923 void (*)(struct super_block *, void *), void *);
2925 extern int dcache_dir_open(struct inode *, struct file *);
2926 extern int dcache_dir_close(struct inode *, struct file *);
2927 extern loff_t dcache_dir_lseek(struct file *, loff_t, int);
2928 extern int dcache_readdir(struct file *, struct dir_context *);
2929 extern int simple_setattr(struct mnt_idmap *, struct dentry *,
2931 extern int simple_getattr(struct mnt_idmap *, const struct path *,
2932 struct kstat *, u32, unsigned int);
2933 extern int simple_statfs(struct dentry *, struct kstatfs *);
2934 extern int simple_open(struct inode *inode, struct file *file);
2935 extern int simple_link(struct dentry *, struct inode *, struct dentry *);
2936 extern int simple_unlink(struct inode *, struct dentry *);
2937 extern int simple_rmdir(struct inode *, struct dentry *);
2938 extern int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry,
2939 struct inode *new_dir, struct dentry *new_dentry);
2940 extern int simple_rename(struct mnt_idmap *, struct inode *,
2941 struct dentry *, struct inode *, struct dentry *,
2943 extern void simple_recursive_removal(struct dentry *,
2944 void (*callback)(struct dentry *));
2945 extern int noop_fsync(struct file *, loff_t, loff_t, int);
2946 extern ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter);
2947 extern int simple_empty(struct dentry *);
2948 extern int simple_write_begin(struct file *file, struct address_space *mapping,
2949 loff_t pos, unsigned len,
2950 struct page **pagep, void **fsdata);
2951 extern const struct address_space_operations ram_aops;
2952 extern int always_delete_dentry(const struct dentry *);
2953 extern struct inode *alloc_anon_inode(struct super_block *);
2954 extern int simple_nosetlease(struct file *, long, struct file_lock **, void **);
2955 extern const struct dentry_operations simple_dentry_operations;
2957 extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags);
2958 extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *);
2959 extern const struct file_operations simple_dir_operations;
2960 extern const struct inode_operations simple_dir_inode_operations;
2961 extern void make_empty_dir_inode(struct inode *inode);
2962 extern bool is_empty_dir_inode(struct inode *inode);
2963 struct tree_descr { const char *name; const struct file_operations *ops; int mode; };
2964 struct dentry *d_alloc_name(struct dentry *, const char *);
2965 extern int simple_fill_super(struct super_block *, unsigned long,
2966 const struct tree_descr *);
2967 extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count);
2968 extern void simple_release_fs(struct vfsmount **mount, int *count);
2970 extern ssize_t simple_read_from_buffer(void __user *to, size_t count,
2971 loff_t *ppos, const void *from, size_t available);
2972 extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
2973 const void __user *from, size_t count);
2975 extern int __generic_file_fsync(struct file *, loff_t, loff_t, int);
2976 extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
2978 extern int generic_check_addressable(unsigned, u64);
2980 extern void generic_set_encrypted_ci_d_ops(struct dentry *dentry);
2982 int may_setattr(struct mnt_idmap *idmap, struct inode *inode,
2983 unsigned int ia_valid);
2984 int setattr_prepare(struct mnt_idmap *, struct dentry *, struct iattr *);
2985 extern int inode_newsize_ok(const struct inode *, loff_t offset);
2986 void setattr_copy(struct mnt_idmap *, struct inode *inode,
2987 const struct iattr *attr);
2989 extern int file_update_time(struct file *file);
2991 static inline bool vma_is_dax(const struct vm_area_struct *vma)
2993 return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host);
2996 static inline bool vma_is_fsdax(struct vm_area_struct *vma)
2998 struct inode *inode;
3000 if (!IS_ENABLED(CONFIG_FS_DAX) || !vma->vm_file)
3002 if (!vma_is_dax(vma))
3004 inode = file_inode(vma->vm_file);
3005 if (S_ISCHR(inode->i_mode))
3006 return false; /* device-dax */
3010 static inline int iocb_flags(struct file *file)
3013 if (file->f_flags & O_APPEND)
3015 if (file->f_flags & O_DIRECT)
3017 if (file->f_flags & O_DSYNC)
3019 if (file->f_flags & __O_SYNC)
3024 static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags)
3026 int kiocb_flags = 0;
3028 /* make sure there's no overlap between RWF and private IOCB flags */
3029 BUILD_BUG_ON((__force int) RWF_SUPPORTED & IOCB_EVENTFD);
3033 if (unlikely(flags & ~RWF_SUPPORTED))
3036 if (flags & RWF_NOWAIT) {
3037 if (!(ki->ki_filp->f_mode & FMODE_NOWAIT))
3039 kiocb_flags |= IOCB_NOIO;
3041 kiocb_flags |= (__force int) (flags & RWF_SUPPORTED);
3042 if (flags & RWF_SYNC)
3043 kiocb_flags |= IOCB_DSYNC;
3045 ki->ki_flags |= kiocb_flags;
3049 static inline ino_t parent_ino(struct dentry *dentry)
3054 * Don't strictly need d_lock here? If the parent ino could change
3055 * then surely we'd have a deeper race in the caller?
3057 spin_lock(&dentry->d_lock);
3058 res = dentry->d_parent->d_inode->i_ino;
3059 spin_unlock(&dentry->d_lock);
3063 /* Transaction based IO helpers */
3066 * An argresp is stored in an allocated page and holds the
3067 * size of the argument or response, along with its content
3069 struct simple_transaction_argresp {
3074 #define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp))
3076 char *simple_transaction_get(struct file *file, const char __user *buf,
3078 ssize_t simple_transaction_read(struct file *file, char __user *buf,
3079 size_t size, loff_t *pos);
3080 int simple_transaction_release(struct inode *inode, struct file *file);
3082 void simple_transaction_set(struct file *file, size_t n);
3085 * simple attribute files
3087 * These attributes behave similar to those in sysfs:
3089 * Writing to an attribute immediately sets a value, an open file can be
3090 * written to multiple times.
3092 * Reading from an attribute creates a buffer from the value that might get
3093 * read with multiple read calls. When the attribute has been read
3094 * completely, no further read calls are possible until the file is opened
3097 * All attributes contain a text representation of a numeric value
3098 * that are accessed with the get() and set() functions.
3100 #define DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, __is_signed) \
3101 static int __fops ## _open(struct inode *inode, struct file *file) \
3103 __simple_attr_check_format(__fmt, 0ull); \
3104 return simple_attr_open(inode, file, __get, __set, __fmt); \
3106 static const struct file_operations __fops = { \
3107 .owner = THIS_MODULE, \
3108 .open = __fops ## _open, \
3109 .release = simple_attr_release, \
3110 .read = simple_attr_read, \
3111 .write = (__is_signed) ? simple_attr_write_signed : simple_attr_write, \
3112 .llseek = generic_file_llseek, \
3115 #define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt) \
3116 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, false)
3118 #define DEFINE_SIMPLE_ATTRIBUTE_SIGNED(__fops, __get, __set, __fmt) \
3119 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, true)
3121 static inline __printf(1, 2)
3122 void __simple_attr_check_format(const char *fmt, ...)
3124 /* don't do anything, just let the compiler check the arguments; */
3127 int simple_attr_open(struct inode *inode, struct file *file,
3128 int (*get)(void *, u64 *), int (*set)(void *, u64),
3130 int simple_attr_release(struct inode *inode, struct file *file);
3131 ssize_t simple_attr_read(struct file *file, char __user *buf,
3132 size_t len, loff_t *ppos);
3133 ssize_t simple_attr_write(struct file *file, const char __user *buf,
3134 size_t len, loff_t *ppos);
3135 ssize_t simple_attr_write_signed(struct file *file, const char __user *buf,
3136 size_t len, loff_t *ppos);
3139 int __init list_bdev_fs_names(char *buf, size_t size);
3141 #define __FMODE_EXEC ((__force int) FMODE_EXEC)
3142 #define __FMODE_NONOTIFY ((__force int) FMODE_NONOTIFY)
3144 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
3145 #define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \
3146 (flag & __FMODE_NONOTIFY)))
3148 static inline bool is_sxid(umode_t mode)
3150 return mode & (S_ISUID | S_ISGID);
3153 static inline int check_sticky(struct mnt_idmap *idmap,
3154 struct inode *dir, struct inode *inode)
3156 if (!(dir->i_mode & S_ISVTX))
3159 return __check_sticky(idmap, dir, inode);
3162 static inline void inode_has_no_xattr(struct inode *inode)
3164 if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC))
3165 inode->i_flags |= S_NOSEC;
3168 static inline bool is_root_inode(struct inode *inode)
3170 return inode == inode->i_sb->s_root->d_inode;
3173 static inline bool dir_emit(struct dir_context *ctx,
3174 const char *name, int namelen,
3175 u64 ino, unsigned type)
3177 return ctx->actor(ctx, name, namelen, ctx->pos, ino, type);
3179 static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx)
3181 return ctx->actor(ctx, ".", 1, ctx->pos,
3182 file->f_path.dentry->d_inode->i_ino, DT_DIR);
3184 static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx)
3186 return ctx->actor(ctx, "..", 2, ctx->pos,
3187 parent_ino(file->f_path.dentry), DT_DIR);
3189 static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx)
3191 if (ctx->pos == 0) {
3192 if (!dir_emit_dot(file, ctx))
3196 if (ctx->pos == 1) {
3197 if (!dir_emit_dotdot(file, ctx))
3203 static inline bool dir_relax(struct inode *inode)
3205 inode_unlock(inode);
3207 return !IS_DEADDIR(inode);
3210 static inline bool dir_relax_shared(struct inode *inode)
3212 inode_unlock_shared(inode);
3213 inode_lock_shared(inode);
3214 return !IS_DEADDIR(inode);
3217 extern bool path_noexec(const struct path *path);
3218 extern void inode_nohighmem(struct inode *inode);
3221 extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len,
3223 extern int generic_fadvise(struct file *file, loff_t offset, loff_t len,
3226 #endif /* _LINUX_FS_H */