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 /* File is opened with O_NDELAY (only set for block devices) */
123 #define FMODE_NDELAY ((__force fmode_t)0x40)
124 /* File is opened with O_EXCL (only set for block devices) */
125 #define FMODE_EXCL ((__force fmode_t)0x80)
126 /* File is opened using open(.., 3, ..) and is writeable only for ioctls
127 (specialy hack for floppy.c) */
128 #define FMODE_WRITE_IOCTL ((__force fmode_t)0x100)
129 /* 32bit hashes as llseek() offset (for directories) */
130 #define FMODE_32BITHASH ((__force fmode_t)0x200)
131 /* 64bit hashes as llseek() offset (for directories) */
132 #define FMODE_64BITHASH ((__force fmode_t)0x400)
135 * Don't update ctime and mtime.
137 * Currently a special hack for the XFS open_by_handle ioctl, but we'll
138 * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon.
140 #define FMODE_NOCMTIME ((__force fmode_t)0x800)
142 /* Expect random access pattern */
143 #define FMODE_RANDOM ((__force fmode_t)0x1000)
145 /* File is huge (eg. /dev/mem): treat loff_t as unsigned */
146 #define FMODE_UNSIGNED_OFFSET ((__force fmode_t)0x2000)
148 /* File is opened with O_PATH; almost nothing can be done with it */
149 #define FMODE_PATH ((__force fmode_t)0x4000)
151 /* File needs atomic accesses to f_pos */
152 #define FMODE_ATOMIC_POS ((__force fmode_t)0x8000)
153 /* Write access to underlying fs */
154 #define FMODE_WRITER ((__force fmode_t)0x10000)
155 /* Has read method(s) */
156 #define FMODE_CAN_READ ((__force fmode_t)0x20000)
157 /* Has write method(s) */
158 #define FMODE_CAN_WRITE ((__force fmode_t)0x40000)
160 #define FMODE_OPENED ((__force fmode_t)0x80000)
161 #define FMODE_CREATED ((__force fmode_t)0x100000)
163 /* File is stream-like */
164 #define FMODE_STREAM ((__force fmode_t)0x200000)
166 /* File supports DIRECT IO */
167 #define FMODE_CAN_ODIRECT ((__force fmode_t)0x400000)
169 /* File was opened by fanotify and shouldn't generate fanotify events */
170 #define FMODE_NONOTIFY ((__force fmode_t)0x4000000)
172 /* File is capable of returning -EAGAIN if I/O will block */
173 #define FMODE_NOWAIT ((__force fmode_t)0x8000000)
175 /* File represents mount that needs unmounting */
176 #define FMODE_NEED_UNMOUNT ((__force fmode_t)0x10000000)
178 /* File does not contribute to nr_files count */
179 #define FMODE_NOACCOUNT ((__force fmode_t)0x20000000)
181 /* File supports async buffered reads */
182 #define FMODE_BUF_RASYNC ((__force fmode_t)0x40000000)
184 /* File supports async nowait buffered writes */
185 #define FMODE_BUF_WASYNC ((__force fmode_t)0x80000000)
188 * Attribute flags. These should be or-ed together to figure out what
191 #define ATTR_MODE (1 << 0)
192 #define ATTR_UID (1 << 1)
193 #define ATTR_GID (1 << 2)
194 #define ATTR_SIZE (1 << 3)
195 #define ATTR_ATIME (1 << 4)
196 #define ATTR_MTIME (1 << 5)
197 #define ATTR_CTIME (1 << 6)
198 #define ATTR_ATIME_SET (1 << 7)
199 #define ATTR_MTIME_SET (1 << 8)
200 #define ATTR_FORCE (1 << 9) /* Not a change, but a change it */
201 #define ATTR_KILL_SUID (1 << 11)
202 #define ATTR_KILL_SGID (1 << 12)
203 #define ATTR_FILE (1 << 13)
204 #define ATTR_KILL_PRIV (1 << 14)
205 #define ATTR_OPEN (1 << 15) /* Truncating from open(O_TRUNC) */
206 #define ATTR_TIMES_SET (1 << 16)
207 #define ATTR_TOUCH (1 << 17)
210 * Whiteout is represented by a char device. The following constants define the
211 * mode and device number to use.
213 #define WHITEOUT_MODE 0
214 #define WHITEOUT_DEV 0
217 * This is the Inode Attributes structure, used for notify_change(). It
218 * uses the above definitions as flags, to know which values have changed.
219 * Also, in this manner, a Filesystem can look at only the values it cares
220 * about. Basically, these are the attributes that the VFS layer can
221 * request to change from the FS layer.
223 * Derek Atkins <warlord@MIT.EDU> 94-10-20
226 unsigned int ia_valid;
229 * The two anonymous unions wrap structures with the same member.
231 * Filesystems raising FS_ALLOW_IDMAP need to use ia_vfs{g,u}id which
232 * are a dedicated type requiring the filesystem to use the dedicated
233 * helpers. Other filesystem can continue to use ia_{g,u}id until they
236 * They always contain the same value. In other words FS_ALLOW_IDMAP
237 * pass down the same value on idmapped mounts as they would on regular
249 struct timespec64 ia_atime;
250 struct timespec64 ia_mtime;
251 struct timespec64 ia_ctime;
254 * Not an attribute, but an auxiliary info for filesystems wanting to
255 * implement an ftruncate() like method. NOTE: filesystem should
256 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL).
258 struct file *ia_file;
262 * Includes for diskquotas.
264 #include <linux/quota.h>
267 * Maximum number of layers of fs stack. Needs to be limited to
268 * prevent kernel stack overflow
270 #define FILESYSTEM_MAX_STACK_DEPTH 2
273 * enum positive_aop_returns - aop return codes with specific semantics
275 * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has
276 * completed, that the page is still locked, and
277 * should be considered active. The VM uses this hint
278 * to return the page to the active list -- it won't
279 * be a candidate for writeback again in the near
280 * future. Other callers must be careful to unlock
281 * the page if they get this return. Returned by
284 * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has
285 * unlocked it and the page might have been truncated.
286 * The caller should back up to acquiring a new page and
287 * trying again. The aop will be taking reasonable
288 * precautions not to livelock. If the caller held a page
289 * reference, it should drop it before retrying. Returned
292 * address_space_operation functions return these large constants to indicate
293 * special semantics to the caller. These are much larger than the bytes in a
294 * page to allow for functions that return the number of bytes operated on in a
298 enum positive_aop_returns {
299 AOP_WRITEPAGE_ACTIVATE = 0x80000,
300 AOP_TRUNCATED_PAGE = 0x80001,
304 * oh the beauties of C type declarations.
307 struct address_space;
308 struct writeback_control;
309 struct readahead_control;
312 * Write life time hint values.
313 * Stored in struct inode as u8.
316 WRITE_LIFE_NOT_SET = 0,
317 WRITE_LIFE_NONE = RWH_WRITE_LIFE_NONE,
318 WRITE_LIFE_SHORT = RWH_WRITE_LIFE_SHORT,
319 WRITE_LIFE_MEDIUM = RWH_WRITE_LIFE_MEDIUM,
320 WRITE_LIFE_LONG = RWH_WRITE_LIFE_LONG,
321 WRITE_LIFE_EXTREME = RWH_WRITE_LIFE_EXTREME,
324 /* Match RWF_* bits to IOCB bits */
325 #define IOCB_HIPRI (__force int) RWF_HIPRI
326 #define IOCB_DSYNC (__force int) RWF_DSYNC
327 #define IOCB_SYNC (__force int) RWF_SYNC
328 #define IOCB_NOWAIT (__force int) RWF_NOWAIT
329 #define IOCB_APPEND (__force int) RWF_APPEND
331 /* non-RWF related bits - start at 16 */
332 #define IOCB_EVENTFD (1 << 16)
333 #define IOCB_DIRECT (1 << 17)
334 #define IOCB_WRITE (1 << 18)
335 /* iocb->ki_waitq is valid */
336 #define IOCB_WAITQ (1 << 19)
337 #define IOCB_NOIO (1 << 20)
338 /* can use bio alloc cache */
339 #define IOCB_ALLOC_CACHE (1 << 21)
342 struct file *ki_filp;
344 void (*ki_complete)(struct kiocb *iocb, long ret);
347 u16 ki_ioprio; /* See linux/ioprio.h */
348 struct wait_page_queue *ki_waitq; /* for async buffered IO */
351 static inline bool is_sync_kiocb(struct kiocb *kiocb)
353 return kiocb->ki_complete == NULL;
356 struct address_space_operations {
357 int (*writepage)(struct page *page, struct writeback_control *wbc);
358 int (*read_folio)(struct file *, struct folio *);
360 /* Write back some dirty pages from this mapping. */
361 int (*writepages)(struct address_space *, struct writeback_control *);
363 /* Mark a folio dirty. Return true if this dirtied it */
364 bool (*dirty_folio)(struct address_space *, struct folio *);
366 void (*readahead)(struct readahead_control *);
368 int (*write_begin)(struct file *, struct address_space *mapping,
369 loff_t pos, unsigned len,
370 struct page **pagep, void **fsdata);
371 int (*write_end)(struct file *, struct address_space *mapping,
372 loff_t pos, unsigned len, unsigned copied,
373 struct page *page, void *fsdata);
375 /* Unfortunately this kludge is needed for FIBMAP. Don't use it */
376 sector_t (*bmap)(struct address_space *, sector_t);
377 void (*invalidate_folio) (struct folio *, size_t offset, size_t len);
378 bool (*release_folio)(struct folio *, gfp_t);
379 void (*free_folio)(struct folio *folio);
380 ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter);
382 * migrate the contents of a folio to the specified target. If
383 * migrate_mode is MIGRATE_ASYNC, it must not block.
385 int (*migrate_folio)(struct address_space *, struct folio *dst,
386 struct folio *src, enum migrate_mode);
387 int (*launder_folio)(struct folio *);
388 bool (*is_partially_uptodate) (struct folio *, size_t from,
390 void (*is_dirty_writeback) (struct folio *, bool *dirty, bool *wb);
391 int (*error_remove_page)(struct address_space *, struct page *);
393 /* swapfile support */
394 int (*swap_activate)(struct swap_info_struct *sis, struct file *file,
396 void (*swap_deactivate)(struct file *file);
397 int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter);
400 extern const struct address_space_operations empty_aops;
403 * struct address_space - Contents of a cacheable, mappable object.
404 * @host: Owner, either the inode or the block_device.
405 * @i_pages: Cached pages.
406 * @invalidate_lock: Guards coherency between page cache contents and
407 * file offset->disk block mappings in the filesystem during invalidates.
408 * It is also used to block modification of page cache contents through
410 * @gfp_mask: Memory allocation flags to use for allocating pages.
411 * @i_mmap_writable: Number of VM_SHARED mappings.
412 * @nr_thps: Number of THPs in the pagecache (non-shmem only).
413 * @i_mmap: Tree of private and shared mappings.
414 * @i_mmap_rwsem: Protects @i_mmap and @i_mmap_writable.
415 * @nrpages: Number of page entries, protected by the i_pages lock.
416 * @writeback_index: Writeback starts here.
418 * @flags: Error bits and flags (AS_*).
419 * @wb_err: The most recent error which has occurred.
420 * @private_lock: For use by the owner of the address_space.
421 * @private_list: For use by the owner of the address_space.
422 * @private_data: For use by the owner of the address_space.
424 struct address_space {
426 struct xarray i_pages;
427 struct rw_semaphore invalidate_lock;
429 atomic_t i_mmap_writable;
430 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
431 /* number of thp, only for non-shmem files */
434 struct rb_root_cached i_mmap;
435 struct rw_semaphore i_mmap_rwsem;
436 unsigned long nrpages;
437 pgoff_t writeback_index;
438 const struct address_space_operations *a_ops;
441 spinlock_t private_lock;
442 struct list_head private_list;
444 } __attribute__((aligned(sizeof(long)))) __randomize_layout;
446 * On most architectures that alignment is already the case; but
447 * must be enforced here for CRIS, to let the least significant bit
448 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON.
451 /* XArray tags, for tagging dirty and writeback pages in the pagecache. */
452 #define PAGECACHE_TAG_DIRTY XA_MARK_0
453 #define PAGECACHE_TAG_WRITEBACK XA_MARK_1
454 #define PAGECACHE_TAG_TOWRITE XA_MARK_2
457 * Returns true if any of the pages in the mapping are marked with the tag.
459 static inline bool mapping_tagged(struct address_space *mapping, xa_mark_t tag)
461 return xa_marked(&mapping->i_pages, tag);
464 static inline void i_mmap_lock_write(struct address_space *mapping)
466 down_write(&mapping->i_mmap_rwsem);
469 static inline int i_mmap_trylock_write(struct address_space *mapping)
471 return down_write_trylock(&mapping->i_mmap_rwsem);
474 static inline void i_mmap_unlock_write(struct address_space *mapping)
476 up_write(&mapping->i_mmap_rwsem);
479 static inline int i_mmap_trylock_read(struct address_space *mapping)
481 return down_read_trylock(&mapping->i_mmap_rwsem);
484 static inline void i_mmap_lock_read(struct address_space *mapping)
486 down_read(&mapping->i_mmap_rwsem);
489 static inline void i_mmap_unlock_read(struct address_space *mapping)
491 up_read(&mapping->i_mmap_rwsem);
494 static inline void i_mmap_assert_locked(struct address_space *mapping)
496 lockdep_assert_held(&mapping->i_mmap_rwsem);
499 static inline void i_mmap_assert_write_locked(struct address_space *mapping)
501 lockdep_assert_held_write(&mapping->i_mmap_rwsem);
505 * Might pages of this file be mapped into userspace?
507 static inline int mapping_mapped(struct address_space *mapping)
509 return !RB_EMPTY_ROOT(&mapping->i_mmap.rb_root);
513 * Might pages of this file have been modified in userspace?
514 * Note that i_mmap_writable counts all VM_SHARED vmas: do_mmap
515 * marks vma as VM_SHARED if it is shared, and the file was opened for
516 * writing i.e. vma may be mprotected writable even if now readonly.
518 * If i_mmap_writable is negative, no new writable mappings are allowed. You
519 * can only deny writable mappings, if none exists right now.
521 static inline int mapping_writably_mapped(struct address_space *mapping)
523 return atomic_read(&mapping->i_mmap_writable) > 0;
526 static inline int mapping_map_writable(struct address_space *mapping)
528 return atomic_inc_unless_negative(&mapping->i_mmap_writable) ?
532 static inline void mapping_unmap_writable(struct address_space *mapping)
534 atomic_dec(&mapping->i_mmap_writable);
537 static inline int mapping_deny_writable(struct address_space *mapping)
539 return atomic_dec_unless_positive(&mapping->i_mmap_writable) ?
543 static inline void mapping_allow_writable(struct address_space *mapping)
545 atomic_inc(&mapping->i_mmap_writable);
549 * Use sequence counter to get consistent i_size on 32-bit processors.
551 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
552 #include <linux/seqlock.h>
553 #define __NEED_I_SIZE_ORDERED
554 #define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount)
556 #define i_size_ordered_init(inode) do { } while (0)
560 #define ACL_NOT_CACHED ((void *)(-1))
562 * ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to
563 * cache the ACL. This also means that ->get_inode_acl() can be called in RCU
564 * mode with the LOOKUP_RCU flag.
566 #define ACL_DONT_CACHE ((void *)(-3))
568 static inline struct posix_acl *
569 uncached_acl_sentinel(struct task_struct *task)
571 return (void *)task + 1;
575 is_uncached_acl(struct posix_acl *acl)
577 return (long)acl & 1;
580 #define IOP_FASTPERM 0x0001
581 #define IOP_LOOKUP 0x0002
582 #define IOP_NOFOLLOW 0x0004
583 #define IOP_XATTR 0x0008
584 #define IOP_DEFAULT_READLINK 0x0010
586 struct fsnotify_mark_connector;
589 * Keep mostly read-only and often accessed (especially for
590 * the RCU path lookup and 'stat' data) fields at the beginning
591 * of the 'struct inode'
595 unsigned short i_opflags;
598 unsigned int i_flags;
600 #ifdef CONFIG_FS_POSIX_ACL
601 struct posix_acl *i_acl;
602 struct posix_acl *i_default_acl;
605 const struct inode_operations *i_op;
606 struct super_block *i_sb;
607 struct address_space *i_mapping;
609 #ifdef CONFIG_SECURITY
613 /* Stat data, not accessed from path walking */
616 * Filesystems may only read i_nlink directly. They shall use the
617 * following functions for modification:
619 * (set|clear|inc|drop)_nlink
620 * inode_(inc|dec)_link_count
623 const unsigned int i_nlink;
624 unsigned int __i_nlink;
628 struct timespec64 i_atime;
629 struct timespec64 i_mtime;
630 struct timespec64 i_ctime;
631 spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */
632 unsigned short i_bytes;
637 #ifdef __NEED_I_SIZE_ORDERED
638 seqcount_t i_size_seqcount;
642 unsigned long i_state;
643 struct rw_semaphore i_rwsem;
645 unsigned long dirtied_when; /* jiffies of first dirtying */
646 unsigned long dirtied_time_when;
648 struct hlist_node i_hash;
649 struct list_head i_io_list; /* backing dev IO list */
650 #ifdef CONFIG_CGROUP_WRITEBACK
651 struct bdi_writeback *i_wb; /* the associated cgroup wb */
653 /* foreign inode detection, see wbc_detach_inode() */
655 u16 i_wb_frn_avg_time;
656 u16 i_wb_frn_history;
658 struct list_head i_lru; /* inode LRU list */
659 struct list_head i_sb_list;
660 struct list_head i_wb_list; /* backing dev writeback list */
662 struct hlist_head i_dentry;
663 struct rcu_head i_rcu;
665 atomic64_t i_version;
666 atomic64_t i_sequence; /* see futex */
668 atomic_t i_dio_count;
669 atomic_t i_writecount;
670 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
671 atomic_t i_readcount; /* struct files open RO */
674 const struct file_operations *i_fop; /* former ->i_op->default_file_ops */
675 void (*free_inode)(struct inode *);
677 struct file_lock_context *i_flctx;
678 struct address_space i_data;
679 struct list_head i_devices;
681 struct pipe_inode_info *i_pipe;
689 #ifdef CONFIG_FSNOTIFY
690 __u32 i_fsnotify_mask; /* all events this inode cares about */
691 struct fsnotify_mark_connector __rcu *i_fsnotify_marks;
694 #ifdef CONFIG_FS_ENCRYPTION
695 struct fscrypt_info *i_crypt_info;
698 #ifdef CONFIG_FS_VERITY
699 struct fsverity_info *i_verity_info;
702 void *i_private; /* fs or device private pointer */
703 } __randomize_layout;
705 struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode);
707 static inline unsigned int i_blocksize(const struct inode *node)
709 return (1 << node->i_blkbits);
712 static inline int inode_unhashed(struct inode *inode)
714 return hlist_unhashed(&inode->i_hash);
718 * __mark_inode_dirty expects inodes to be hashed. Since we don't
719 * want special inodes in the fileset inode space, we make them
720 * appear hashed, but do not put on any lists. hlist_del()
721 * will work fine and require no locking.
723 static inline void inode_fake_hash(struct inode *inode)
725 hlist_add_fake(&inode->i_hash);
729 * inode->i_mutex nesting subclasses for the lock validator:
731 * 0: the object of the current VFS operation
735 * 4: second non-directory
736 * 5: second parent (when locking independent directories in rename)
738 * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two
739 * non-directories at once.
741 * The locking order between these classes is
742 * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory
744 enum inode_i_mutex_lock_class
754 static inline void inode_lock(struct inode *inode)
756 down_write(&inode->i_rwsem);
759 static inline void inode_unlock(struct inode *inode)
761 up_write(&inode->i_rwsem);
764 static inline void inode_lock_shared(struct inode *inode)
766 down_read(&inode->i_rwsem);
769 static inline void inode_unlock_shared(struct inode *inode)
771 up_read(&inode->i_rwsem);
774 static inline int inode_trylock(struct inode *inode)
776 return down_write_trylock(&inode->i_rwsem);
779 static inline int inode_trylock_shared(struct inode *inode)
781 return down_read_trylock(&inode->i_rwsem);
784 static inline int inode_is_locked(struct inode *inode)
786 return rwsem_is_locked(&inode->i_rwsem);
789 static inline void inode_lock_nested(struct inode *inode, unsigned subclass)
791 down_write_nested(&inode->i_rwsem, subclass);
794 static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass)
796 down_read_nested(&inode->i_rwsem, subclass);
799 static inline void filemap_invalidate_lock(struct address_space *mapping)
801 down_write(&mapping->invalidate_lock);
804 static inline void filemap_invalidate_unlock(struct address_space *mapping)
806 up_write(&mapping->invalidate_lock);
809 static inline void filemap_invalidate_lock_shared(struct address_space *mapping)
811 down_read(&mapping->invalidate_lock);
814 static inline int filemap_invalidate_trylock_shared(
815 struct address_space *mapping)
817 return down_read_trylock(&mapping->invalidate_lock);
820 static inline void filemap_invalidate_unlock_shared(
821 struct address_space *mapping)
823 up_read(&mapping->invalidate_lock);
826 void lock_two_nondirectories(struct inode *, struct inode*);
827 void unlock_two_nondirectories(struct inode *, struct inode*);
829 void filemap_invalidate_lock_two(struct address_space *mapping1,
830 struct address_space *mapping2);
831 void filemap_invalidate_unlock_two(struct address_space *mapping1,
832 struct address_space *mapping2);
836 * NOTE: in a 32bit arch with a preemptable kernel and
837 * an UP compile the i_size_read/write must be atomic
838 * with respect to the local cpu (unlike with preempt disabled),
839 * but they don't need to be atomic with respect to other cpus like in
840 * true SMP (so they need either to either locally disable irq around
841 * the read or for example on x86 they can be still implemented as a
842 * cmpxchg8b without the need of the lock prefix). For SMP compiles
843 * and 64bit archs it makes no difference if preempt is enabled or not.
845 static inline loff_t i_size_read(const struct inode *inode)
847 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
852 seq = read_seqcount_begin(&inode->i_size_seqcount);
853 i_size = inode->i_size;
854 } while (read_seqcount_retry(&inode->i_size_seqcount, seq));
856 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
860 i_size = inode->i_size;
864 return inode->i_size;
869 * NOTE: unlike i_size_read(), i_size_write() does need locking around it
870 * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount
871 * can be lost, resulting in subsequent i_size_read() calls spinning forever.
873 static inline void i_size_write(struct inode *inode, loff_t i_size)
875 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
877 write_seqcount_begin(&inode->i_size_seqcount);
878 inode->i_size = i_size;
879 write_seqcount_end(&inode->i_size_seqcount);
881 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
883 inode->i_size = i_size;
886 inode->i_size = i_size;
890 static inline unsigned iminor(const struct inode *inode)
892 return MINOR(inode->i_rdev);
895 static inline unsigned imajor(const struct inode *inode)
897 return MAJOR(inode->i_rdev);
901 rwlock_t lock; /* protects pid, uid, euid fields */
902 struct pid *pid; /* pid or -pgrp where SIGIO should be sent */
903 enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */
904 kuid_t uid, euid; /* uid/euid of process setting the owner */
905 int signum; /* posix.1b rt signal to be delivered on IO */
909 * struct file_ra_state - Track a file's readahead state.
910 * @start: Where the most recent readahead started.
911 * @size: Number of pages read in the most recent readahead.
912 * @async_size: Numer of pages that were/are not needed immediately
913 * and so were/are genuinely "ahead". Start next readahead when
914 * the first of these pages is accessed.
915 * @ra_pages: Maximum size of a readahead request, copied from the bdi.
916 * @mmap_miss: How many mmap accesses missed in the page cache.
917 * @prev_pos: The last byte in the most recent read request.
919 * When this structure is passed to ->readahead(), the "most recent"
920 * readahead means the current readahead.
922 struct file_ra_state {
925 unsigned int async_size;
926 unsigned int ra_pages;
927 unsigned int mmap_miss;
932 * Check if @index falls in the readahead windows.
934 static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index)
936 return (index >= ra->start &&
937 index < ra->start + ra->size);
942 struct llist_node f_llist;
943 struct rcu_head f_rcuhead;
944 unsigned int f_iocb_flags;
947 struct inode *f_inode; /* cached value */
948 const struct file_operations *f_op;
951 * Protects f_ep, f_flags.
952 * Must not be taken from IRQ context.
955 atomic_long_t f_count;
956 unsigned int f_flags;
958 struct mutex f_pos_lock;
960 struct fown_struct f_owner;
961 const struct cred *f_cred;
962 struct file_ra_state f_ra;
965 #ifdef CONFIG_SECURITY
968 /* needed for tty driver, and maybe others */
972 /* Used by fs/eventpoll.c to link all the hooks to this file */
973 struct hlist_head *f_ep;
974 #endif /* #ifdef CONFIG_EPOLL */
975 struct address_space *f_mapping;
977 errseq_t f_sb_err; /* for syncfs */
979 __attribute__((aligned(4))); /* lest something weird decides that 2 is OK */
984 /* file identifier */
985 unsigned char f_handle[];
988 static inline struct file *get_file(struct file *f)
990 atomic_long_inc(&f->f_count);
993 #define get_file_rcu(x) atomic_long_inc_not_zero(&(x)->f_count)
994 #define file_count(x) atomic_long_read(&(x)->f_count)
996 #define MAX_NON_LFS ((1UL<<31) - 1)
998 /* Page cache limit. The filesystems should put that into their s_maxbytes
999 limits, otherwise bad things can happen in VM. */
1000 #if BITS_PER_LONG==32
1001 #define MAX_LFS_FILESIZE ((loff_t)ULONG_MAX << PAGE_SHIFT)
1002 #elif BITS_PER_LONG==64
1003 #define MAX_LFS_FILESIZE ((loff_t)LLONG_MAX)
1006 /* legacy typedef, should eventually be removed */
1007 typedef void *fl_owner_t;
1011 /* The following constant reflects the upper bound of the file/locking space */
1013 #define OFFSET_MAX type_max(loff_t)
1014 #define OFFT_OFFSET_MAX type_max(off_t)
1017 extern void send_sigio(struct fown_struct *fown, int fd, int band);
1019 static inline struct inode *file_inode(const struct file *f)
1024 static inline struct dentry *file_dentry(const struct file *file)
1026 return d_real(file->f_path.dentry, file_inode(file));
1029 struct fasync_struct {
1033 struct fasync_struct *fa_next; /* singly linked list */
1034 struct file *fa_file;
1035 struct rcu_head fa_rcu;
1038 #define FASYNC_MAGIC 0x4601
1040 /* SMP safe fasync helpers: */
1041 extern int fasync_helper(int, struct file *, int, struct fasync_struct **);
1042 extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *);
1043 extern int fasync_remove_entry(struct file *, struct fasync_struct **);
1044 extern struct fasync_struct *fasync_alloc(void);
1045 extern void fasync_free(struct fasync_struct *);
1047 /* can be called from interrupts */
1048 extern void kill_fasync(struct fasync_struct **, int, int);
1050 extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force);
1051 extern int f_setown(struct file *filp, unsigned long arg, int force);
1052 extern void f_delown(struct file *filp);
1053 extern pid_t f_getown(struct file *filp);
1054 extern int send_sigurg(struct fown_struct *fown);
1057 * sb->s_flags. Note that these mirror the equivalent MS_* flags where
1058 * represented in both.
1060 #define SB_RDONLY 1 /* Mount read-only */
1061 #define SB_NOSUID 2 /* Ignore suid and sgid bits */
1062 #define SB_NODEV 4 /* Disallow access to device special files */
1063 #define SB_NOEXEC 8 /* Disallow program execution */
1064 #define SB_SYNCHRONOUS 16 /* Writes are synced at once */
1065 #define SB_MANDLOCK 64 /* Allow mandatory locks on an FS */
1066 #define SB_DIRSYNC 128 /* Directory modifications are synchronous */
1067 #define SB_NOATIME 1024 /* Do not update access times. */
1068 #define SB_NODIRATIME 2048 /* Do not update directory access times */
1069 #define SB_SILENT 32768
1070 #define SB_POSIXACL (1<<16) /* VFS does not apply the umask */
1071 #define SB_INLINECRYPT (1<<17) /* Use blk-crypto for encrypted files */
1072 #define SB_KERNMOUNT (1<<22) /* this is a kern_mount call */
1073 #define SB_I_VERSION (1<<23) /* Update inode I_version field */
1074 #define SB_LAZYTIME (1<<25) /* Update the on-disk [acm]times lazily */
1076 /* These sb flags are internal to the kernel */
1077 #define SB_SUBMOUNT (1<<26)
1078 #define SB_FORCE (1<<27)
1079 #define SB_NOSEC (1<<28)
1080 #define SB_BORN (1<<29)
1081 #define SB_ACTIVE (1<<30)
1082 #define SB_NOUSER (1<<31)
1084 /* These flags relate to encoding and casefolding */
1085 #define SB_ENC_STRICT_MODE_FL (1 << 0)
1087 #define sb_has_strict_encoding(sb) \
1088 (sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL)
1094 #define MNT_FORCE 0x00000001 /* Attempt to forcibily umount */
1095 #define MNT_DETACH 0x00000002 /* Just detach from the tree */
1096 #define MNT_EXPIRE 0x00000004 /* Mark for expiry */
1097 #define UMOUNT_NOFOLLOW 0x00000008 /* Don't follow symlink on umount */
1098 #define UMOUNT_UNUSED 0x80000000 /* Flag guaranteed to be unused */
1101 #define SB_I_CGROUPWB 0x00000001 /* cgroup-aware writeback enabled */
1102 #define SB_I_NOEXEC 0x00000002 /* Ignore executables on this fs */
1103 #define SB_I_NODEV 0x00000004 /* Ignore devices on this fs */
1104 #define SB_I_STABLE_WRITES 0x00000008 /* don't modify blks until WB is done */
1106 /* sb->s_iflags to limit user namespace mounts */
1107 #define SB_I_USERNS_VISIBLE 0x00000010 /* fstype already mounted */
1108 #define SB_I_IMA_UNVERIFIABLE_SIGNATURE 0x00000020
1109 #define SB_I_UNTRUSTED_MOUNTER 0x00000040
1111 #define SB_I_SKIP_SYNC 0x00000100 /* Skip superblock at global sync */
1112 #define SB_I_PERSB_BDI 0x00000200 /* has a per-sb bdi */
1113 #define SB_I_TS_EXPIRY_WARNED 0x00000400 /* warned about timestamp range expiry */
1114 #define SB_I_RETIRED 0x00000800 /* superblock shouldn't be reused */
1116 /* Possible states of 'frozen' field */
1118 SB_UNFROZEN = 0, /* FS is unfrozen */
1119 SB_FREEZE_WRITE = 1, /* Writes, dir ops, ioctls frozen */
1120 SB_FREEZE_PAGEFAULT = 2, /* Page faults stopped as well */
1121 SB_FREEZE_FS = 3, /* For internal FS use (e.g. to stop
1122 * internal threads if needed) */
1123 SB_FREEZE_COMPLETE = 4, /* ->freeze_fs finished successfully */
1126 #define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1)
1129 int frozen; /* Is sb frozen? */
1130 wait_queue_head_t wait_unfrozen; /* wait for thaw */
1131 struct percpu_rw_semaphore rw_sem[SB_FREEZE_LEVELS];
1134 struct super_block {
1135 struct list_head s_list; /* Keep this first */
1136 dev_t s_dev; /* search index; _not_ kdev_t */
1137 unsigned char s_blocksize_bits;
1138 unsigned long s_blocksize;
1139 loff_t s_maxbytes; /* Max file size */
1140 struct file_system_type *s_type;
1141 const struct super_operations *s_op;
1142 const struct dquot_operations *dq_op;
1143 const struct quotactl_ops *s_qcop;
1144 const struct export_operations *s_export_op;
1145 unsigned long s_flags;
1146 unsigned long s_iflags; /* internal SB_I_* flags */
1147 unsigned long s_magic;
1148 struct dentry *s_root;
1149 struct rw_semaphore s_umount;
1152 #ifdef CONFIG_SECURITY
1155 const struct xattr_handler **s_xattr;
1156 #ifdef CONFIG_FS_ENCRYPTION
1157 const struct fscrypt_operations *s_cop;
1158 struct fscrypt_keyring *s_master_keys; /* master crypto keys in use */
1160 #ifdef CONFIG_FS_VERITY
1161 const struct fsverity_operations *s_vop;
1163 #if IS_ENABLED(CONFIG_UNICODE)
1164 struct unicode_map *s_encoding;
1165 __u16 s_encoding_flags;
1167 struct hlist_bl_head s_roots; /* alternate root dentries for NFS */
1168 struct list_head s_mounts; /* list of mounts; _not_ for fs use */
1169 struct block_device *s_bdev;
1170 struct backing_dev_info *s_bdi;
1171 struct mtd_info *s_mtd;
1172 struct hlist_node s_instances;
1173 unsigned int s_quota_types; /* Bitmask of supported quota types */
1174 struct quota_info s_dquot; /* Diskquota specific options */
1176 struct sb_writers s_writers;
1179 * Keep s_fs_info, s_time_gran, s_fsnotify_mask, and
1180 * s_fsnotify_marks together for cache efficiency. They are frequently
1181 * accessed and rarely modified.
1183 void *s_fs_info; /* Filesystem private info */
1185 /* Granularity of c/m/atime in ns (cannot be worse than a second) */
1187 /* Time limits for c/m/atime in seconds */
1188 time64_t s_time_min;
1189 time64_t s_time_max;
1190 #ifdef CONFIG_FSNOTIFY
1191 __u32 s_fsnotify_mask;
1192 struct fsnotify_mark_connector __rcu *s_fsnotify_marks;
1195 char s_id[32]; /* Informational name */
1196 uuid_t s_uuid; /* UUID */
1198 unsigned int s_max_links;
1202 * The next field is for VFS *only*. No filesystems have any business
1203 * even looking at it. You had been warned.
1205 struct mutex s_vfs_rename_mutex; /* Kludge */
1208 * Filesystem subtype. If non-empty the filesystem type field
1209 * in /proc/mounts will be "type.subtype"
1211 const char *s_subtype;
1213 const struct dentry_operations *s_d_op; /* default d_op for dentries */
1215 struct shrinker s_shrink; /* per-sb shrinker handle */
1217 /* Number of inodes with nlink == 0 but still referenced */
1218 atomic_long_t s_remove_count;
1221 * Number of inode/mount/sb objects that are being watched, note that
1222 * inodes objects are currently double-accounted.
1224 atomic_long_t s_fsnotify_connectors;
1226 /* Being remounted read-only */
1227 int s_readonly_remount;
1229 /* per-sb errseq_t for reporting writeback errors via syncfs */
1232 /* AIO completions deferred from interrupt context */
1233 struct workqueue_struct *s_dio_done_wq;
1234 struct hlist_head s_pins;
1237 * Owning user namespace and default context in which to
1238 * interpret filesystem uids, gids, quotas, device nodes,
1239 * xattrs and security labels.
1241 struct user_namespace *s_user_ns;
1244 * The list_lru structure is essentially just a pointer to a table
1245 * of per-node lru lists, each of which has its own spinlock.
1246 * There is no need to put them into separate cachelines.
1248 struct list_lru s_dentry_lru;
1249 struct list_lru s_inode_lru;
1250 struct rcu_head rcu;
1251 struct work_struct destroy_work;
1253 struct mutex s_sync_lock; /* sync serialisation lock */
1256 * Indicates how deep in a filesystem stack this SB is
1260 /* s_inode_list_lock protects s_inodes */
1261 spinlock_t s_inode_list_lock ____cacheline_aligned_in_smp;
1262 struct list_head s_inodes; /* all inodes */
1264 spinlock_t s_inode_wblist_lock;
1265 struct list_head s_inodes_wb; /* writeback inodes */
1266 } __randomize_layout;
1268 static inline struct user_namespace *i_user_ns(const struct inode *inode)
1270 return inode->i_sb->s_user_ns;
1273 /* Helper functions so that in most cases filesystems will
1274 * not need to deal directly with kuid_t and kgid_t and can
1275 * instead deal with the raw numeric values that are stored
1276 * in the filesystem.
1278 static inline uid_t i_uid_read(const struct inode *inode)
1280 return from_kuid(i_user_ns(inode), inode->i_uid);
1283 static inline gid_t i_gid_read(const struct inode *inode)
1285 return from_kgid(i_user_ns(inode), inode->i_gid);
1288 static inline void i_uid_write(struct inode *inode, uid_t uid)
1290 inode->i_uid = make_kuid(i_user_ns(inode), uid);
1293 static inline void i_gid_write(struct inode *inode, gid_t gid)
1295 inode->i_gid = make_kgid(i_user_ns(inode), gid);
1299 * i_uid_into_vfsuid - map an inode's i_uid down according to an idmapping
1300 * @idmap: idmap of the mount the inode was found from
1301 * @inode: inode to map
1303 * Return: whe inode's i_uid mapped down according to @idmap.
1304 * If the inode's i_uid has no mapping INVALID_VFSUID is returned.
1306 static inline vfsuid_t i_uid_into_vfsuid(struct mnt_idmap *idmap,
1307 const struct inode *inode)
1309 return make_vfsuid(idmap, i_user_ns(inode), inode->i_uid);
1313 * i_uid_needs_update - check whether inode's i_uid needs to be updated
1314 * @idmap: idmap of the mount the inode was found from
1315 * @attr: the new attributes of @inode
1316 * @inode: the inode to update
1318 * Check whether the $inode's i_uid field needs to be updated taking idmapped
1319 * mounts into account if the filesystem supports it.
1321 * Return: true if @inode's i_uid field needs to be updated, false if not.
1323 static inline bool i_uid_needs_update(struct mnt_idmap *idmap,
1324 const struct iattr *attr,
1325 const struct inode *inode)
1327 return ((attr->ia_valid & ATTR_UID) &&
1328 !vfsuid_eq(attr->ia_vfsuid,
1329 i_uid_into_vfsuid(idmap, inode)));
1333 * i_uid_update - update @inode's i_uid field
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 * Safely update @inode's i_uid field translating the vfsuid of any idmapped
1339 * mount into the filesystem kuid.
1341 static inline void i_uid_update(struct mnt_idmap *idmap,
1342 const struct iattr *attr,
1343 struct inode *inode)
1345 if (attr->ia_valid & ATTR_UID)
1346 inode->i_uid = from_vfsuid(idmap, i_user_ns(inode),
1351 * i_gid_into_vfsgid - map an inode's i_gid down according to an idmapping
1352 * @idmap: idmap of the mount the inode was found from
1353 * @inode: inode to map
1355 * Return: the inode's i_gid mapped down according to @idmap.
1356 * If the inode's i_gid has no mapping INVALID_VFSGID is returned.
1358 static inline vfsgid_t i_gid_into_vfsgid(struct mnt_idmap *idmap,
1359 const struct inode *inode)
1361 return make_vfsgid(idmap, i_user_ns(inode), inode->i_gid);
1365 * i_gid_needs_update - check whether inode's i_gid needs to be updated
1366 * @idmap: idmap of the mount the inode was found from
1367 * @attr: the new attributes of @inode
1368 * @inode: the inode to update
1370 * Check whether the $inode's i_gid field needs to be updated taking idmapped
1371 * mounts into account if the filesystem supports it.
1373 * Return: true if @inode's i_gid field needs to be updated, false if not.
1375 static inline bool i_gid_needs_update(struct mnt_idmap *idmap,
1376 const struct iattr *attr,
1377 const struct inode *inode)
1379 return ((attr->ia_valid & ATTR_GID) &&
1380 !vfsgid_eq(attr->ia_vfsgid,
1381 i_gid_into_vfsgid(idmap, inode)));
1385 * i_gid_update - update @inode's i_gid field
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 * Safely update @inode's i_gid field translating the vfsgid of any idmapped
1391 * mount into the filesystem kgid.
1393 static inline void i_gid_update(struct mnt_idmap *idmap,
1394 const struct iattr *attr,
1395 struct inode *inode)
1397 if (attr->ia_valid & ATTR_GID)
1398 inode->i_gid = from_vfsgid(idmap, i_user_ns(inode),
1403 * inode_fsuid_set - initialize inode's i_uid field with callers fsuid
1404 * @inode: inode to initialize
1405 * @idmap: idmap of the mount the inode was found from
1407 * Initialize the i_uid field of @inode. If the inode was found/created via
1408 * an idmapped mount map the caller's fsuid according to @idmap.
1410 static inline void inode_fsuid_set(struct inode *inode,
1411 struct mnt_idmap *idmap)
1413 inode->i_uid = mapped_fsuid(idmap, i_user_ns(inode));
1417 * inode_fsgid_set - initialize inode's i_gid field with callers fsgid
1418 * @inode: inode to initialize
1419 * @idmap: idmap of the mount the inode was found from
1421 * Initialize the i_gid field of @inode. If the inode was found/created via
1422 * an idmapped mount map the caller's fsgid according to @idmap.
1424 static inline void inode_fsgid_set(struct inode *inode,
1425 struct mnt_idmap *idmap)
1427 inode->i_gid = mapped_fsgid(idmap, i_user_ns(inode));
1431 * fsuidgid_has_mapping() - check whether caller's fsuid/fsgid is mapped
1432 * @sb: the superblock we want a mapping in
1433 * @idmap: idmap of the relevant mount
1435 * Check whether the caller's fsuid and fsgid have a valid mapping in the
1436 * s_user_ns of the superblock @sb. If the caller is on an idmapped mount map
1437 * the caller's fsuid and fsgid according to the @idmap first.
1439 * Return: true if fsuid and fsgid is mapped, false if not.
1441 static inline bool fsuidgid_has_mapping(struct super_block *sb,
1442 struct mnt_idmap *idmap)
1444 struct user_namespace *fs_userns = sb->s_user_ns;
1448 kuid = mapped_fsuid(idmap, fs_userns);
1449 if (!uid_valid(kuid))
1451 kgid = mapped_fsgid(idmap, fs_userns);
1452 if (!gid_valid(kgid))
1454 return kuid_has_mapping(fs_userns, kuid) &&
1455 kgid_has_mapping(fs_userns, kgid);
1458 extern struct timespec64 current_time(struct inode *inode);
1461 * Snapshotting support.
1465 * These are internal functions, please use sb_start_{write,pagefault,intwrite}
1468 static inline void __sb_end_write(struct super_block *sb, int level)
1470 percpu_up_read(sb->s_writers.rw_sem + level-1);
1473 static inline void __sb_start_write(struct super_block *sb, int level)
1475 percpu_down_read(sb->s_writers.rw_sem + level - 1);
1478 static inline bool __sb_start_write_trylock(struct super_block *sb, int level)
1480 return percpu_down_read_trylock(sb->s_writers.rw_sem + level - 1);
1483 #define __sb_writers_acquired(sb, lev) \
1484 percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1485 #define __sb_writers_release(sb, lev) \
1486 percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1488 static inline bool sb_write_started(const struct super_block *sb)
1490 return lockdep_is_held_type(sb->s_writers.rw_sem + SB_FREEZE_WRITE - 1, 1);
1494 * sb_end_write - drop write access to a superblock
1495 * @sb: the super we wrote to
1497 * Decrement number of writers to the filesystem. Wake up possible waiters
1498 * wanting to freeze the filesystem.
1500 static inline void sb_end_write(struct super_block *sb)
1502 __sb_end_write(sb, SB_FREEZE_WRITE);
1506 * sb_end_pagefault - drop write access to a superblock from a page fault
1507 * @sb: the super we wrote to
1509 * Decrement number of processes handling write page fault to the filesystem.
1510 * Wake up possible waiters wanting to freeze the filesystem.
1512 static inline void sb_end_pagefault(struct super_block *sb)
1514 __sb_end_write(sb, SB_FREEZE_PAGEFAULT);
1518 * sb_end_intwrite - drop write access to a superblock for internal fs purposes
1519 * @sb: the super we wrote to
1521 * Decrement fs-internal number of writers to the filesystem. Wake up possible
1522 * waiters wanting to freeze the filesystem.
1524 static inline void sb_end_intwrite(struct super_block *sb)
1526 __sb_end_write(sb, SB_FREEZE_FS);
1530 * sb_start_write - get write access to a superblock
1531 * @sb: the super we write to
1533 * When a process wants to write data or metadata to a file system (i.e. dirty
1534 * a page or an inode), it should embed the operation in a sb_start_write() -
1535 * sb_end_write() pair to get exclusion against file system freezing. This
1536 * function increments number of writers preventing freezing. If the file
1537 * system is already frozen, the function waits until the file system is
1540 * Since freeze protection behaves as a lock, users have to preserve
1541 * ordering of freeze protection and other filesystem locks. Generally,
1542 * freeze protection should be the outermost lock. In particular, we have:
1545 * -> i_mutex (write path, truncate, directory ops, ...)
1546 * -> s_umount (freeze_super, thaw_super)
1548 static inline void sb_start_write(struct super_block *sb)
1550 __sb_start_write(sb, SB_FREEZE_WRITE);
1553 static inline bool sb_start_write_trylock(struct super_block *sb)
1555 return __sb_start_write_trylock(sb, SB_FREEZE_WRITE);
1559 * sb_start_pagefault - get write access to a superblock from a page fault
1560 * @sb: the super we write to
1562 * When a process starts handling write page fault, it should embed the
1563 * operation into sb_start_pagefault() - sb_end_pagefault() pair to get
1564 * exclusion against file system freezing. This is needed since the page fault
1565 * is going to dirty a page. This function increments number of running page
1566 * faults preventing freezing. If the file system is already frozen, the
1567 * function waits until the file system is thawed.
1569 * Since page fault freeze protection behaves as a lock, users have to preserve
1570 * ordering of freeze protection and other filesystem locks. It is advised to
1571 * put sb_start_pagefault() close to mmap_lock in lock ordering. Page fault
1572 * handling code implies lock dependency:
1575 * -> sb_start_pagefault
1577 static inline void sb_start_pagefault(struct super_block *sb)
1579 __sb_start_write(sb, SB_FREEZE_PAGEFAULT);
1583 * sb_start_intwrite - get write access to a superblock for internal fs purposes
1584 * @sb: the super we write to
1586 * This is the third level of protection against filesystem freezing. It is
1587 * free for use by a filesystem. The only requirement is that it must rank
1588 * below sb_start_pagefault.
1590 * For example filesystem can call sb_start_intwrite() when starting a
1591 * transaction which somewhat eases handling of freezing for internal sources
1592 * of filesystem changes (internal fs threads, discarding preallocation on file
1595 static inline void sb_start_intwrite(struct super_block *sb)
1597 __sb_start_write(sb, SB_FREEZE_FS);
1600 static inline bool sb_start_intwrite_trylock(struct super_block *sb)
1602 return __sb_start_write_trylock(sb, SB_FREEZE_FS);
1605 bool inode_owner_or_capable(struct mnt_idmap *idmap,
1606 const struct inode *inode);
1609 * VFS helper functions..
1611 int vfs_create(struct mnt_idmap *, struct inode *,
1612 struct dentry *, umode_t, bool);
1613 int vfs_mkdir(struct mnt_idmap *, struct inode *,
1614 struct dentry *, umode_t);
1615 int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *,
1617 int vfs_symlink(struct mnt_idmap *, struct inode *,
1618 struct dentry *, const char *);
1619 int vfs_link(struct dentry *, struct mnt_idmap *, struct inode *,
1620 struct dentry *, struct inode **);
1621 int vfs_rmdir(struct mnt_idmap *, struct inode *, struct dentry *);
1622 int vfs_unlink(struct mnt_idmap *, struct inode *, struct dentry *,
1626 * struct renamedata - contains all information required for renaming
1627 * @old_mnt_idmap: idmap of the old mount the inode was found from
1628 * @old_dir: parent of source
1629 * @old_dentry: source
1630 * @new_mnt_idmap: idmap of the new mount the inode was found from
1631 * @new_dir: parent of destination
1632 * @new_dentry: destination
1633 * @delegated_inode: returns an inode needing a delegation break
1634 * @flags: rename flags
1637 struct mnt_idmap *old_mnt_idmap;
1638 struct inode *old_dir;
1639 struct dentry *old_dentry;
1640 struct mnt_idmap *new_mnt_idmap;
1641 struct inode *new_dir;
1642 struct dentry *new_dentry;
1643 struct inode **delegated_inode;
1645 } __randomize_layout;
1647 int vfs_rename(struct renamedata *);
1649 static inline int vfs_whiteout(struct mnt_idmap *idmap,
1650 struct inode *dir, struct dentry *dentry)
1652 return vfs_mknod(idmap, dir, dentry, S_IFCHR | WHITEOUT_MODE,
1656 struct file *vfs_tmpfile_open(struct mnt_idmap *idmap,
1657 const struct path *parentpath,
1658 umode_t mode, int open_flag, const struct cred *cred);
1660 int vfs_mkobj(struct dentry *, umode_t,
1661 int (*f)(struct dentry *, umode_t, void *),
1664 int vfs_fchown(struct file *file, uid_t user, gid_t group);
1665 int vfs_fchmod(struct file *file, umode_t mode);
1666 int vfs_utimes(const struct path *path, struct timespec64 *times);
1668 extern long vfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1670 #ifdef CONFIG_COMPAT
1671 extern long compat_ptr_ioctl(struct file *file, unsigned int cmd,
1674 #define compat_ptr_ioctl NULL
1678 * VFS file helper functions.
1680 void inode_init_owner(struct mnt_idmap *idmap, struct inode *inode,
1681 const struct inode *dir, umode_t mode);
1682 extern bool may_open_dev(const struct path *path);
1683 umode_t mode_strip_sgid(struct mnt_idmap *idmap,
1684 const struct inode *dir, umode_t mode);
1687 * This is the "filldir" function type, used by readdir() to let
1688 * the kernel specify what kind of dirent layout it wants to have.
1689 * This allows the kernel to read directories into kernel space or
1690 * to have different dirent layouts depending on the binary type.
1691 * Return 'true' to keep going and 'false' if there are no more entries.
1694 typedef bool (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64,
1697 struct dir_context {
1703 * These flags let !MMU mmap() govern direct device mapping vs immediate
1704 * copying more easily for MAP_PRIVATE, especially for ROM filesystems.
1706 * NOMMU_MAP_COPY: Copy can be mapped (MAP_PRIVATE)
1707 * NOMMU_MAP_DIRECT: Can be mapped directly (MAP_SHARED)
1708 * NOMMU_MAP_READ: Can be mapped for reading
1709 * NOMMU_MAP_WRITE: Can be mapped for writing
1710 * NOMMU_MAP_EXEC: Can be mapped for execution
1712 #define NOMMU_MAP_COPY 0x00000001
1713 #define NOMMU_MAP_DIRECT 0x00000008
1714 #define NOMMU_MAP_READ VM_MAYREAD
1715 #define NOMMU_MAP_WRITE VM_MAYWRITE
1716 #define NOMMU_MAP_EXEC VM_MAYEXEC
1718 #define NOMMU_VMFLAGS \
1719 (NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC)
1722 * These flags control the behavior of the remap_file_range function pointer.
1723 * If it is called with len == 0 that means "remap to end of source file".
1724 * See Documentation/filesystems/vfs.rst for more details about this call.
1726 * REMAP_FILE_DEDUP: only remap if contents identical (i.e. deduplicate)
1727 * REMAP_FILE_CAN_SHORTEN: caller can handle a shortened request
1729 #define REMAP_FILE_DEDUP (1 << 0)
1730 #define REMAP_FILE_CAN_SHORTEN (1 << 1)
1733 * These flags signal that the caller is ok with altering various aspects of
1734 * the behavior of the remap operation. The changes must be made by the
1735 * implementation; the vfs remap helper functions can take advantage of them.
1736 * Flags in this category exist to preserve the quirky behavior of the hoisted
1737 * btrfs clone/dedupe ioctls.
1739 #define REMAP_FILE_ADVISORY (REMAP_FILE_CAN_SHORTEN)
1742 * These flags control the behavior of vfs_copy_file_range().
1743 * They are not available to the user via syscall.
1745 * COPY_FILE_SPLICE: call splice direct instead of fs clone/copy ops
1747 #define COPY_FILE_SPLICE (1 << 0)
1750 struct io_uring_cmd;
1752 struct file_operations {
1753 struct module *owner;
1754 loff_t (*llseek) (struct file *, loff_t, int);
1755 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
1756 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
1757 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
1758 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
1759 int (*iopoll)(struct kiocb *kiocb, struct io_comp_batch *,
1760 unsigned int flags);
1761 int (*iterate) (struct file *, struct dir_context *);
1762 int (*iterate_shared) (struct file *, struct dir_context *);
1763 __poll_t (*poll) (struct file *, struct poll_table_struct *);
1764 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
1765 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
1766 int (*mmap) (struct file *, struct vm_area_struct *);
1767 unsigned long mmap_supported_flags;
1768 int (*open) (struct inode *, struct file *);
1769 int (*flush) (struct file *, fl_owner_t id);
1770 int (*release) (struct inode *, struct file *);
1771 int (*fsync) (struct file *, loff_t, loff_t, int datasync);
1772 int (*fasync) (int, struct file *, int);
1773 int (*lock) (struct file *, int, struct file_lock *);
1774 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
1775 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1776 int (*check_flags)(int);
1777 int (*flock) (struct file *, int, struct file_lock *);
1778 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
1779 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
1780 int (*setlease)(struct file *, long, struct file_lock **, void **);
1781 long (*fallocate)(struct file *file, int mode, loff_t offset,
1783 void (*show_fdinfo)(struct seq_file *m, struct file *f);
1785 unsigned (*mmap_capabilities)(struct file *);
1787 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *,
1788 loff_t, size_t, unsigned int);
1789 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in,
1790 struct file *file_out, loff_t pos_out,
1791 loff_t len, unsigned int remap_flags);
1792 int (*fadvise)(struct file *, loff_t, loff_t, int);
1793 int (*uring_cmd)(struct io_uring_cmd *ioucmd, unsigned int issue_flags);
1794 int (*uring_cmd_iopoll)(struct io_uring_cmd *, struct io_comp_batch *,
1795 unsigned int poll_flags);
1796 } __randomize_layout;
1798 struct inode_operations {
1799 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
1800 const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *);
1801 int (*permission) (struct mnt_idmap *, struct inode *, int);
1802 struct posix_acl * (*get_inode_acl)(struct inode *, int, bool);
1804 int (*readlink) (struct dentry *, char __user *,int);
1806 int (*create) (struct mnt_idmap *, struct inode *,struct dentry *,
1808 int (*link) (struct dentry *,struct inode *,struct dentry *);
1809 int (*unlink) (struct inode *,struct dentry *);
1810 int (*symlink) (struct mnt_idmap *, struct inode *,struct dentry *,
1812 int (*mkdir) (struct mnt_idmap *, struct inode *,struct dentry *,
1814 int (*rmdir) (struct inode *,struct dentry *);
1815 int (*mknod) (struct mnt_idmap *, struct inode *,struct dentry *,
1817 int (*rename) (struct mnt_idmap *, struct inode *, struct dentry *,
1818 struct inode *, struct dentry *, unsigned int);
1819 int (*setattr) (struct mnt_idmap *, struct dentry *, struct iattr *);
1820 int (*getattr) (struct mnt_idmap *, const struct path *,
1821 struct kstat *, u32, unsigned int);
1822 ssize_t (*listxattr) (struct dentry *, char *, size_t);
1823 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start,
1825 int (*update_time)(struct inode *, struct timespec64 *, int);
1826 int (*atomic_open)(struct inode *, struct dentry *,
1827 struct file *, unsigned open_flag,
1828 umode_t create_mode);
1829 int (*tmpfile) (struct mnt_idmap *, struct inode *,
1830 struct file *, umode_t);
1831 struct posix_acl *(*get_acl)(struct mnt_idmap *, struct dentry *,
1833 int (*set_acl)(struct mnt_idmap *, struct dentry *,
1834 struct posix_acl *, int);
1835 int (*fileattr_set)(struct mnt_idmap *idmap,
1836 struct dentry *dentry, struct fileattr *fa);
1837 int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa);
1838 } ____cacheline_aligned;
1840 static inline ssize_t call_read_iter(struct file *file, struct kiocb *kio,
1841 struct iov_iter *iter)
1843 return file->f_op->read_iter(kio, iter);
1846 static inline ssize_t call_write_iter(struct file *file, struct kiocb *kio,
1847 struct iov_iter *iter)
1849 return file->f_op->write_iter(kio, iter);
1852 static inline int call_mmap(struct file *file, struct vm_area_struct *vma)
1854 return file->f_op->mmap(file, vma);
1857 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
1858 extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *);
1859 extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *,
1860 loff_t, size_t, unsigned int);
1861 extern ssize_t generic_copy_file_range(struct file *file_in, loff_t pos_in,
1862 struct file *file_out, loff_t pos_out,
1863 size_t len, unsigned int flags);
1864 int __generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
1865 struct file *file_out, loff_t pos_out,
1866 loff_t *len, unsigned int remap_flags,
1867 const struct iomap_ops *dax_read_ops);
1868 int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
1869 struct file *file_out, loff_t pos_out,
1870 loff_t *count, unsigned int remap_flags);
1871 extern loff_t do_clone_file_range(struct file *file_in, loff_t pos_in,
1872 struct file *file_out, loff_t pos_out,
1873 loff_t len, unsigned int remap_flags);
1874 extern loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in,
1875 struct file *file_out, loff_t pos_out,
1876 loff_t len, unsigned int remap_flags);
1877 extern int vfs_dedupe_file_range(struct file *file,
1878 struct file_dedupe_range *same);
1879 extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos,
1880 struct file *dst_file, loff_t dst_pos,
1881 loff_t len, unsigned int remap_flags);
1884 struct super_operations {
1885 struct inode *(*alloc_inode)(struct super_block *sb);
1886 void (*destroy_inode)(struct inode *);
1887 void (*free_inode)(struct inode *);
1889 void (*dirty_inode) (struct inode *, int flags);
1890 int (*write_inode) (struct inode *, struct writeback_control *wbc);
1891 int (*drop_inode) (struct inode *);
1892 void (*evict_inode) (struct inode *);
1893 void (*put_super) (struct super_block *);
1894 int (*sync_fs)(struct super_block *sb, int wait);
1895 int (*freeze_super) (struct super_block *);
1896 int (*freeze_fs) (struct super_block *);
1897 int (*thaw_super) (struct super_block *);
1898 int (*unfreeze_fs) (struct super_block *);
1899 int (*statfs) (struct dentry *, struct kstatfs *);
1900 int (*remount_fs) (struct super_block *, int *, char *);
1901 void (*umount_begin) (struct super_block *);
1903 int (*show_options)(struct seq_file *, struct dentry *);
1904 int (*show_devname)(struct seq_file *, struct dentry *);
1905 int (*show_path)(struct seq_file *, struct dentry *);
1906 int (*show_stats)(struct seq_file *, struct dentry *);
1908 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
1909 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
1910 struct dquot **(*get_dquots)(struct inode *);
1912 long (*nr_cached_objects)(struct super_block *,
1913 struct shrink_control *);
1914 long (*free_cached_objects)(struct super_block *,
1915 struct shrink_control *);
1919 * Inode flags - they have no relation to superblock flags now
1921 #define S_SYNC (1 << 0) /* Writes are synced at once */
1922 #define S_NOATIME (1 << 1) /* Do not update access times */
1923 #define S_APPEND (1 << 2) /* Append-only file */
1924 #define S_IMMUTABLE (1 << 3) /* Immutable file */
1925 #define S_DEAD (1 << 4) /* removed, but still open directory */
1926 #define S_NOQUOTA (1 << 5) /* Inode is not counted to quota */
1927 #define S_DIRSYNC (1 << 6) /* Directory modifications are synchronous */
1928 #define S_NOCMTIME (1 << 7) /* Do not update file c/mtime */
1929 #define S_SWAPFILE (1 << 8) /* Do not truncate: swapon got its bmaps */
1930 #define S_PRIVATE (1 << 9) /* Inode is fs-internal */
1931 #define S_IMA (1 << 10) /* Inode has an associated IMA struct */
1932 #define S_AUTOMOUNT (1 << 11) /* Automount/referral quasi-directory */
1933 #define S_NOSEC (1 << 12) /* no suid or xattr security attributes */
1934 #ifdef CONFIG_FS_DAX
1935 #define S_DAX (1 << 13) /* Direct Access, avoiding the page cache */
1937 #define S_DAX 0 /* Make all the DAX code disappear */
1939 #define S_ENCRYPTED (1 << 14) /* Encrypted file (using fs/crypto/) */
1940 #define S_CASEFOLD (1 << 15) /* Casefolded file */
1941 #define S_VERITY (1 << 16) /* Verity file (using fs/verity/) */
1942 #define S_KERNEL_FILE (1 << 17) /* File is in use by the kernel (eg. fs/cachefiles) */
1945 * Note that nosuid etc flags are inode-specific: setting some file-system
1946 * flags just means all the inodes inherit those flags by default. It might be
1947 * possible to override it selectively if you really wanted to with some
1948 * ioctl() that is not currently implemented.
1950 * Exception: SB_RDONLY is always applied to the entire file system.
1952 * Unfortunately, it is possible to change a filesystems flags with it mounted
1953 * with files in use. This means that all of the inodes will not have their
1954 * i_flags updated. Hence, i_flags no longer inherit the superblock mount
1955 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org
1957 #define __IS_FLG(inode, flg) ((inode)->i_sb->s_flags & (flg))
1959 static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; }
1960 #define IS_RDONLY(inode) sb_rdonly((inode)->i_sb)
1961 #define IS_SYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS) || \
1962 ((inode)->i_flags & S_SYNC))
1963 #define IS_DIRSYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS|SB_DIRSYNC) || \
1964 ((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
1965 #define IS_MANDLOCK(inode) __IS_FLG(inode, SB_MANDLOCK)
1966 #define IS_NOATIME(inode) __IS_FLG(inode, SB_RDONLY|SB_NOATIME)
1967 #define IS_I_VERSION(inode) __IS_FLG(inode, SB_I_VERSION)
1969 #define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA)
1970 #define IS_APPEND(inode) ((inode)->i_flags & S_APPEND)
1971 #define IS_IMMUTABLE(inode) ((inode)->i_flags & S_IMMUTABLE)
1972 #define IS_POSIXACL(inode) __IS_FLG(inode, SB_POSIXACL)
1974 #define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD)
1975 #define IS_NOCMTIME(inode) ((inode)->i_flags & S_NOCMTIME)
1976 #define IS_SWAPFILE(inode) ((inode)->i_flags & S_SWAPFILE)
1977 #define IS_PRIVATE(inode) ((inode)->i_flags & S_PRIVATE)
1978 #define IS_IMA(inode) ((inode)->i_flags & S_IMA)
1979 #define IS_AUTOMOUNT(inode) ((inode)->i_flags & S_AUTOMOUNT)
1980 #define IS_NOSEC(inode) ((inode)->i_flags & S_NOSEC)
1981 #define IS_DAX(inode) ((inode)->i_flags & S_DAX)
1982 #define IS_ENCRYPTED(inode) ((inode)->i_flags & S_ENCRYPTED)
1983 #define IS_CASEFOLDED(inode) ((inode)->i_flags & S_CASEFOLD)
1984 #define IS_VERITY(inode) ((inode)->i_flags & S_VERITY)
1986 #define IS_WHITEOUT(inode) (S_ISCHR(inode->i_mode) && \
1987 (inode)->i_rdev == WHITEOUT_DEV)
1989 static inline bool HAS_UNMAPPED_ID(struct mnt_idmap *idmap,
1990 struct inode *inode)
1992 return !vfsuid_valid(i_uid_into_vfsuid(idmap, inode)) ||
1993 !vfsgid_valid(i_gid_into_vfsgid(idmap, inode));
1996 static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp)
1998 *kiocb = (struct kiocb) {
2000 .ki_flags = filp->f_iocb_flags,
2001 .ki_ioprio = get_current_ioprio(),
2005 static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
2008 *kiocb = (struct kiocb) {
2010 .ki_flags = kiocb_src->ki_flags,
2011 .ki_ioprio = kiocb_src->ki_ioprio,
2012 .ki_pos = kiocb_src->ki_pos,
2017 * Inode state bits. Protected by inode->i_lock
2019 * Four bits determine the dirty state of the inode: I_DIRTY_SYNC,
2020 * I_DIRTY_DATASYNC, I_DIRTY_PAGES, and I_DIRTY_TIME.
2022 * Four bits define the lifetime of an inode. Initially, inodes are I_NEW,
2023 * until that flag is cleared. I_WILL_FREE, I_FREEING and I_CLEAR are set at
2024 * various stages of removing an inode.
2026 * Two bits are used for locking and completion notification, I_NEW and I_SYNC.
2028 * I_DIRTY_SYNC Inode is dirty, but doesn't have to be written on
2029 * fdatasync() (unless I_DIRTY_DATASYNC is also set).
2030 * Timestamp updates are the usual cause.
2031 * I_DIRTY_DATASYNC Data-related inode changes pending. We keep track of
2032 * these changes separately from I_DIRTY_SYNC so that we
2033 * don't have to write inode on fdatasync() when only
2034 * e.g. the timestamps have changed.
2035 * I_DIRTY_PAGES Inode has dirty pages. Inode itself may be clean.
2036 * I_DIRTY_TIME The inode itself has dirty timestamps, and the
2037 * lazytime mount option is enabled. We keep track of this
2038 * separately from I_DIRTY_SYNC in order to implement
2039 * lazytime. This gets cleared if I_DIRTY_INODE
2040 * (I_DIRTY_SYNC and/or I_DIRTY_DATASYNC) gets set. But
2041 * I_DIRTY_TIME can still be set if I_DIRTY_SYNC is already
2042 * in place because writeback might already be in progress
2043 * and we don't want to lose the time update
2044 * I_NEW Serves as both a mutex and completion notification.
2045 * New inodes set I_NEW. If two processes both create
2046 * the same inode, one of them will release its inode and
2047 * wait for I_NEW to be released before returning.
2048 * Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can
2049 * also cause waiting on I_NEW, without I_NEW actually
2050 * being set. find_inode() uses this to prevent returning
2051 * nearly-dead inodes.
2052 * I_WILL_FREE Must be set when calling write_inode_now() if i_count
2053 * is zero. I_FREEING must be set when I_WILL_FREE is
2055 * I_FREEING Set when inode is about to be freed but still has dirty
2056 * pages or buffers attached or the inode itself is still
2058 * I_CLEAR Added by clear_inode(). In this state the inode is
2059 * clean and can be destroyed. Inode keeps I_FREEING.
2061 * Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are
2062 * prohibited for many purposes. iget() must wait for
2063 * the inode to be completely released, then create it
2064 * anew. Other functions will just ignore such inodes,
2065 * if appropriate. I_NEW is used for waiting.
2067 * I_SYNC Writeback of inode is running. The bit is set during
2068 * data writeback, and cleared with a wakeup on the bit
2069 * address once it is done. The bit is also used to pin
2070 * the inode in memory for flusher thread.
2072 * I_REFERENCED Marks the inode as recently references on the LRU list.
2074 * I_DIO_WAKEUP Never set. Only used as a key for wait_on_bit().
2076 * I_WB_SWITCH Cgroup bdi_writeback switching in progress. Used to
2077 * synchronize competing switching instances and to tell
2078 * wb stat updates to grab the i_pages lock. See
2079 * inode_switch_wbs_work_fn() for details.
2081 * I_OVL_INUSE Used by overlayfs to get exclusive ownership on upper
2082 * and work dirs among overlayfs mounts.
2084 * I_CREATING New object's inode in the middle of setting up.
2086 * I_DONTCACHE Evict inode as soon as it is not used anymore.
2088 * I_SYNC_QUEUED Inode is queued in b_io or b_more_io writeback lists.
2089 * Used to detect that mark_inode_dirty() should not move
2090 * inode between dirty lists.
2092 * I_PINNING_FSCACHE_WB Inode is pinning an fscache object for writeback.
2094 * Q: What is the difference between I_WILL_FREE and I_FREEING?
2096 #define I_DIRTY_SYNC (1 << 0)
2097 #define I_DIRTY_DATASYNC (1 << 1)
2098 #define I_DIRTY_PAGES (1 << 2)
2100 #define I_NEW (1 << __I_NEW)
2101 #define I_WILL_FREE (1 << 4)
2102 #define I_FREEING (1 << 5)
2103 #define I_CLEAR (1 << 6)
2105 #define I_SYNC (1 << __I_SYNC)
2106 #define I_REFERENCED (1 << 8)
2107 #define __I_DIO_WAKEUP 9
2108 #define I_DIO_WAKEUP (1 << __I_DIO_WAKEUP)
2109 #define I_LINKABLE (1 << 10)
2110 #define I_DIRTY_TIME (1 << 11)
2111 #define I_WB_SWITCH (1 << 13)
2112 #define I_OVL_INUSE (1 << 14)
2113 #define I_CREATING (1 << 15)
2114 #define I_DONTCACHE (1 << 16)
2115 #define I_SYNC_QUEUED (1 << 17)
2116 #define I_PINNING_FSCACHE_WB (1 << 18)
2118 #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC)
2119 #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES)
2120 #define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME)
2122 extern void __mark_inode_dirty(struct inode *, int);
2123 static inline void mark_inode_dirty(struct inode *inode)
2125 __mark_inode_dirty(inode, I_DIRTY);
2128 static inline void mark_inode_dirty_sync(struct inode *inode)
2130 __mark_inode_dirty(inode, I_DIRTY_SYNC);
2134 * Returns true if the given inode itself only has dirty timestamps (its pages
2135 * may still be dirty) and isn't currently being allocated or freed.
2136 * Filesystems should call this if when writing an inode when lazytime is
2137 * enabled, they want to opportunistically write the timestamps of other inodes
2138 * located very nearby on-disk, e.g. in the same inode block. This returns true
2139 * if the given inode is in need of such an opportunistic update. Requires
2140 * i_lock, or at least later re-checking under i_lock.
2142 static inline bool inode_is_dirtytime_only(struct inode *inode)
2144 return (inode->i_state & (I_DIRTY_TIME | I_NEW |
2145 I_FREEING | I_WILL_FREE)) == I_DIRTY_TIME;
2148 extern void inc_nlink(struct inode *inode);
2149 extern void drop_nlink(struct inode *inode);
2150 extern void clear_nlink(struct inode *inode);
2151 extern void set_nlink(struct inode *inode, unsigned int nlink);
2153 static inline void inode_inc_link_count(struct inode *inode)
2156 mark_inode_dirty(inode);
2159 static inline void inode_dec_link_count(struct inode *inode)
2162 mark_inode_dirty(inode);
2165 enum file_time_flags {
2172 extern bool atime_needs_update(const struct path *, struct inode *);
2173 extern void touch_atime(const struct path *);
2174 int inode_update_time(struct inode *inode, struct timespec64 *time, int flags);
2176 static inline void file_accessed(struct file *file)
2178 if (!(file->f_flags & O_NOATIME))
2179 touch_atime(&file->f_path);
2182 extern int file_modified(struct file *file);
2183 int kiocb_modified(struct kiocb *iocb);
2185 int sync_inode_metadata(struct inode *inode, int wait);
2187 struct file_system_type {
2190 #define FS_REQUIRES_DEV 1
2191 #define FS_BINARY_MOUNTDATA 2
2192 #define FS_HAS_SUBTYPE 4
2193 #define FS_USERNS_MOUNT 8 /* Can be mounted by userns root */
2194 #define FS_DISALLOW_NOTIFY_PERM 16 /* Disable fanotify permission events */
2195 #define FS_ALLOW_IDMAP 32 /* FS has been updated to handle vfs idmappings. */
2196 #define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() during rename() internally. */
2197 int (*init_fs_context)(struct fs_context *);
2198 const struct fs_parameter_spec *parameters;
2199 struct dentry *(*mount) (struct file_system_type *, int,
2200 const char *, void *);
2201 void (*kill_sb) (struct super_block *);
2202 struct module *owner;
2203 struct file_system_type * next;
2204 struct hlist_head fs_supers;
2206 struct lock_class_key s_lock_key;
2207 struct lock_class_key s_umount_key;
2208 struct lock_class_key s_vfs_rename_key;
2209 struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
2211 struct lock_class_key i_lock_key;
2212 struct lock_class_key i_mutex_key;
2213 struct lock_class_key invalidate_lock_key;
2214 struct lock_class_key i_mutex_dir_key;
2217 #define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME)
2219 extern struct dentry *mount_bdev(struct file_system_type *fs_type,
2220 int flags, const char *dev_name, void *data,
2221 int (*fill_super)(struct super_block *, void *, int));
2222 extern struct dentry *mount_single(struct file_system_type *fs_type,
2223 int flags, void *data,
2224 int (*fill_super)(struct super_block *, void *, int));
2225 extern struct dentry *mount_nodev(struct file_system_type *fs_type,
2226 int flags, void *data,
2227 int (*fill_super)(struct super_block *, void *, int));
2228 extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path);
2229 void retire_super(struct super_block *sb);
2230 void generic_shutdown_super(struct super_block *sb);
2231 void kill_block_super(struct super_block *sb);
2232 void kill_anon_super(struct super_block *sb);
2233 void kill_litter_super(struct super_block *sb);
2234 void deactivate_super(struct super_block *sb);
2235 void deactivate_locked_super(struct super_block *sb);
2236 int set_anon_super(struct super_block *s, void *data);
2237 int set_anon_super_fc(struct super_block *s, struct fs_context *fc);
2238 int get_anon_bdev(dev_t *);
2239 void free_anon_bdev(dev_t);
2240 struct super_block *sget_fc(struct fs_context *fc,
2241 int (*test)(struct super_block *, struct fs_context *),
2242 int (*set)(struct super_block *, struct fs_context *));
2243 struct super_block *sget(struct file_system_type *type,
2244 int (*test)(struct super_block *,void *),
2245 int (*set)(struct super_block *,void *),
2246 int flags, void *data);
2248 /* Alas, no aliases. Too much hassle with bringing module.h everywhere */
2249 #define fops_get(fops) \
2250 (((fops) && try_module_get((fops)->owner) ? (fops) : NULL))
2251 #define fops_put(fops) \
2252 do { if (fops) module_put((fops)->owner); } while(0)
2254 * This one is to be used *ONLY* from ->open() instances.
2255 * fops must be non-NULL, pinned down *and* module dependencies
2256 * should be sufficient to pin the caller down as well.
2258 #define replace_fops(f, fops) \
2260 struct file *__file = (f); \
2261 fops_put(__file->f_op); \
2262 BUG_ON(!(__file->f_op = (fops))); \
2265 extern int register_filesystem(struct file_system_type *);
2266 extern int unregister_filesystem(struct file_system_type *);
2267 extern int vfs_statfs(const struct path *, struct kstatfs *);
2268 extern int user_statfs(const char __user *, struct kstatfs *);
2269 extern int fd_statfs(int, struct kstatfs *);
2270 extern int freeze_super(struct super_block *super);
2271 extern int thaw_super(struct super_block *super);
2272 extern __printf(2, 3)
2273 int super_setup_bdi_name(struct super_block *sb, char *fmt, ...);
2274 extern int super_setup_bdi(struct super_block *sb);
2276 extern int current_umask(void);
2278 extern void ihold(struct inode * inode);
2279 extern void iput(struct inode *);
2280 extern int generic_update_time(struct inode *, struct timespec64 *, int);
2283 extern struct kobject *fs_kobj;
2285 #define MAX_RW_COUNT (INT_MAX & PAGE_MASK)
2290 const char *name; /* pointer to actual string */
2291 const __user char *uptr; /* original userland pointer */
2293 struct audit_names *aname;
2296 static_assert(offsetof(struct filename, iname) % sizeof(long) == 0);
2298 static inline struct mnt_idmap *file_mnt_idmap(struct file *file)
2300 return mnt_idmap(file->f_path.mnt);
2304 * is_idmapped_mnt - check whether a mount is mapped
2305 * @mnt: the mount to check
2307 * If @mnt has an non @nop_mnt_idmap attached to it then @mnt is mapped.
2309 * Return: true if mount is mapped, false if not.
2311 static inline bool is_idmapped_mnt(const struct vfsmount *mnt)
2313 return mnt_idmap(mnt) != &nop_mnt_idmap;
2316 extern long vfs_truncate(const struct path *, loff_t);
2317 int do_truncate(struct mnt_idmap *, struct dentry *, loff_t start,
2318 unsigned int time_attrs, struct file *filp);
2319 extern int vfs_fallocate(struct file *file, int mode, loff_t offset,
2321 extern long do_sys_open(int dfd, const char __user *filename, int flags,
2323 extern struct file *file_open_name(struct filename *, int, umode_t);
2324 extern struct file *filp_open(const char *, int, umode_t);
2325 extern struct file *file_open_root(const struct path *,
2326 const char *, int, umode_t);
2327 static inline struct file *file_open_root_mnt(struct vfsmount *mnt,
2328 const char *name, int flags, umode_t mode)
2330 return file_open_root(&(struct path){.mnt = mnt, .dentry = mnt->mnt_root},
2333 extern struct file * dentry_open(const struct path *, int, const struct cred *);
2334 extern struct file *dentry_create(const struct path *path, int flags,
2335 umode_t mode, const struct cred *cred);
2336 extern struct file * open_with_fake_path(const struct path *, int,
2337 struct inode*, const struct cred *);
2338 static inline struct file *file_clone_open(struct file *file)
2340 return dentry_open(&file->f_path, file->f_flags, file->f_cred);
2342 extern int filp_close(struct file *, fl_owner_t id);
2344 extern struct filename *getname_flags(const char __user *, int, int *);
2345 extern struct filename *getname_uflags(const char __user *, int);
2346 extern struct filename *getname(const char __user *);
2347 extern struct filename *getname_kernel(const char *);
2348 extern void putname(struct filename *name);
2350 extern int finish_open(struct file *file, struct dentry *dentry,
2351 int (*open)(struct inode *, struct file *));
2352 extern int finish_no_open(struct file *file, struct dentry *dentry);
2354 /* Helper for the simple case when original dentry is used */
2355 static inline int finish_open_simple(struct file *file, int error)
2360 return finish_open(file, file->f_path.dentry, NULL);
2364 extern void __init vfs_caches_init_early(void);
2365 extern void __init vfs_caches_init(void);
2367 extern struct kmem_cache *names_cachep;
2369 #define __getname() kmem_cache_alloc(names_cachep, GFP_KERNEL)
2370 #define __putname(name) kmem_cache_free(names_cachep, (void *)(name))
2372 extern struct super_block *blockdev_superblock;
2373 static inline bool sb_is_blkdev_sb(struct super_block *sb)
2375 return IS_ENABLED(CONFIG_BLOCK) && sb == blockdev_superblock;
2378 void emergency_thaw_all(void);
2379 extern int sync_filesystem(struct super_block *);
2380 extern const struct file_operations def_blk_fops;
2381 extern const struct file_operations def_chr_fops;
2384 #define CHRDEV_MAJOR_MAX 512
2385 /* Marks the bottom of the first segment of free char majors */
2386 #define CHRDEV_MAJOR_DYN_END 234
2387 /* Marks the top and bottom of the second segment of free char majors */
2388 #define CHRDEV_MAJOR_DYN_EXT_START 511
2389 #define CHRDEV_MAJOR_DYN_EXT_END 384
2391 extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *);
2392 extern int register_chrdev_region(dev_t, unsigned, const char *);
2393 extern int __register_chrdev(unsigned int major, unsigned int baseminor,
2394 unsigned int count, const char *name,
2395 const struct file_operations *fops);
2396 extern void __unregister_chrdev(unsigned int major, unsigned int baseminor,
2397 unsigned int count, const char *name);
2398 extern void unregister_chrdev_region(dev_t, unsigned);
2399 extern void chrdev_show(struct seq_file *,off_t);
2401 static inline int register_chrdev(unsigned int major, const char *name,
2402 const struct file_operations *fops)
2404 return __register_chrdev(major, 0, 256, name, fops);
2407 static inline void unregister_chrdev(unsigned int major, const char *name)
2409 __unregister_chrdev(major, 0, 256, name);
2412 extern void init_special_inode(struct inode *, umode_t, dev_t);
2414 /* Invalid inode operations -- fs/bad_inode.c */
2415 extern void make_bad_inode(struct inode *);
2416 extern bool is_bad_inode(struct inode *);
2418 extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart,
2420 extern int __must_check file_check_and_advance_wb_err(struct file *file);
2421 extern int __must_check file_write_and_wait_range(struct file *file,
2422 loff_t start, loff_t end);
2424 static inline int file_write_and_wait(struct file *file)
2426 return file_write_and_wait_range(file, 0, LLONG_MAX);
2429 extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end,
2431 extern int vfs_fsync(struct file *file, int datasync);
2433 extern int sync_file_range(struct file *file, loff_t offset, loff_t nbytes,
2434 unsigned int flags);
2436 static inline bool iocb_is_dsync(const struct kiocb *iocb)
2438 return (iocb->ki_flags & IOCB_DSYNC) ||
2439 IS_SYNC(iocb->ki_filp->f_mapping->host);
2443 * Sync the bytes written if this was a synchronous write. Expect ki_pos
2444 * to already be updated for the write, and will return either the amount
2445 * of bytes passed in, or an error if syncing the file failed.
2447 static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count)
2449 if (iocb_is_dsync(iocb)) {
2450 int ret = vfs_fsync_range(iocb->ki_filp,
2451 iocb->ki_pos - count, iocb->ki_pos - 1,
2452 (iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
2460 extern void emergency_sync(void);
2461 extern void emergency_remount(void);
2464 extern int bmap(struct inode *inode, sector_t *block);
2466 static inline int bmap(struct inode *inode, sector_t *block)
2472 int notify_change(struct mnt_idmap *, struct dentry *,
2473 struct iattr *, struct inode **);
2474 int inode_permission(struct mnt_idmap *, struct inode *, int);
2475 int generic_permission(struct mnt_idmap *, struct inode *, int);
2476 static inline int file_permission(struct file *file, int mask)
2478 return inode_permission(file_mnt_idmap(file),
2479 file_inode(file), mask);
2481 static inline int path_permission(const struct path *path, int mask)
2483 return inode_permission(mnt_idmap(path->mnt),
2484 d_inode(path->dentry), mask);
2486 int __check_sticky(struct mnt_idmap *idmap, struct inode *dir,
2487 struct inode *inode);
2489 static inline bool execute_ok(struct inode *inode)
2491 return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode);
2494 static inline bool inode_wrong_type(const struct inode *inode, umode_t mode)
2496 return (inode->i_mode ^ mode) & S_IFMT;
2499 static inline void file_start_write(struct file *file)
2501 if (!S_ISREG(file_inode(file)->i_mode))
2503 sb_start_write(file_inode(file)->i_sb);
2506 static inline bool file_start_write_trylock(struct file *file)
2508 if (!S_ISREG(file_inode(file)->i_mode))
2510 return sb_start_write_trylock(file_inode(file)->i_sb);
2513 static inline void file_end_write(struct file *file)
2515 if (!S_ISREG(file_inode(file)->i_mode))
2517 __sb_end_write(file_inode(file)->i_sb, SB_FREEZE_WRITE);
2521 * This is used for regular files where some users -- especially the
2522 * currently executed binary in a process, previously handled via
2523 * VM_DENYWRITE -- cannot handle concurrent write (and maybe mmap
2524 * read-write shared) accesses.
2526 * get_write_access() gets write permission for a file.
2527 * put_write_access() releases this write permission.
2528 * deny_write_access() denies write access to a file.
2529 * allow_write_access() re-enables write access to a file.
2531 * The i_writecount field of an inode can have the following values:
2532 * 0: no write access, no denied write access
2533 * < 0: (-i_writecount) users that denied write access to the file.
2534 * > 0: (i_writecount) users that have write access to the file.
2536 * Normally we operate on that counter with atomic_{inc,dec} and it's safe
2537 * except for the cases where we don't hold i_writecount yet. Then we need to
2538 * use {get,deny}_write_access() - these functions check the sign and refuse
2539 * to do the change if sign is wrong.
2541 static inline int get_write_access(struct inode *inode)
2543 return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY;
2545 static inline int deny_write_access(struct file *file)
2547 struct inode *inode = file_inode(file);
2548 return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY;
2550 static inline void put_write_access(struct inode * inode)
2552 atomic_dec(&inode->i_writecount);
2554 static inline void allow_write_access(struct file *file)
2557 atomic_inc(&file_inode(file)->i_writecount);
2559 static inline bool inode_is_open_for_write(const struct inode *inode)
2561 return atomic_read(&inode->i_writecount) > 0;
2564 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
2565 static inline void i_readcount_dec(struct inode *inode)
2567 BUG_ON(!atomic_read(&inode->i_readcount));
2568 atomic_dec(&inode->i_readcount);
2570 static inline void i_readcount_inc(struct inode *inode)
2572 atomic_inc(&inode->i_readcount);
2575 static inline void i_readcount_dec(struct inode *inode)
2579 static inline void i_readcount_inc(struct inode *inode)
2584 extern int do_pipe_flags(int *, int);
2586 extern ssize_t kernel_read(struct file *, void *, size_t, loff_t *);
2587 ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos);
2588 extern ssize_t kernel_write(struct file *, const void *, size_t, loff_t *);
2589 extern ssize_t __kernel_write(struct file *, const void *, size_t, loff_t *);
2590 extern struct file * open_exec(const char *);
2592 /* fs/dcache.c -- generic fs support functions */
2593 extern bool is_subdir(struct dentry *, struct dentry *);
2594 extern bool path_is_under(const struct path *, const struct path *);
2596 extern char *file_path(struct file *, char *, int);
2598 #include <linux/err.h>
2600 /* needed for stackable file system support */
2601 extern loff_t default_llseek(struct file *file, loff_t offset, int whence);
2603 extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence);
2605 extern int inode_init_always(struct super_block *, struct inode *);
2606 extern void inode_init_once(struct inode *);
2607 extern void address_space_init_once(struct address_space *mapping);
2608 extern struct inode * igrab(struct inode *);
2609 extern ino_t iunique(struct super_block *, ino_t);
2610 extern int inode_needs_sync(struct inode *inode);
2611 extern int generic_delete_inode(struct inode *inode);
2612 static inline int generic_drop_inode(struct inode *inode)
2614 return !inode->i_nlink || inode_unhashed(inode);
2616 extern void d_mark_dontcache(struct inode *inode);
2618 extern struct inode *ilookup5_nowait(struct super_block *sb,
2619 unsigned long hashval, int (*test)(struct inode *, void *),
2621 extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
2622 int (*test)(struct inode *, void *), void *data);
2623 extern struct inode *ilookup(struct super_block *sb, unsigned long ino);
2625 extern struct inode *inode_insert5(struct inode *inode, unsigned long hashval,
2626 int (*test)(struct inode *, void *),
2627 int (*set)(struct inode *, void *),
2629 extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *);
2630 extern struct inode * iget_locked(struct super_block *, unsigned long);
2631 extern struct inode *find_inode_nowait(struct super_block *,
2633 int (*match)(struct inode *,
2634 unsigned long, void *),
2636 extern struct inode *find_inode_rcu(struct super_block *, unsigned long,
2637 int (*)(struct inode *, void *), void *);
2638 extern struct inode *find_inode_by_ino_rcu(struct super_block *, unsigned long);
2639 extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *);
2640 extern int insert_inode_locked(struct inode *);
2641 #ifdef CONFIG_DEBUG_LOCK_ALLOC
2642 extern void lockdep_annotate_inode_mutex_key(struct inode *inode);
2644 static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { };
2646 extern void unlock_new_inode(struct inode *);
2647 extern void discard_new_inode(struct inode *);
2648 extern unsigned int get_next_ino(void);
2649 extern void evict_inodes(struct super_block *sb);
2650 void dump_mapping(const struct address_space *);
2653 * Userspace may rely on the the inode number being non-zero. For example, glibc
2654 * simply ignores files with zero i_ino in unlink() and other places.
2656 * As an additional complication, if userspace was compiled with
2657 * _FILE_OFFSET_BITS=32 on a 64-bit kernel we'll only end up reading out the
2658 * lower 32 bits, so we need to check that those aren't zero explicitly. With
2659 * _FILE_OFFSET_BITS=64, this may cause some harmless false-negatives, but
2660 * better safe than sorry.
2662 static inline bool is_zero_ino(ino_t ino)
2664 return (u32)ino == 0;
2667 extern void __iget(struct inode * inode);
2668 extern void iget_failed(struct inode *);
2669 extern void clear_inode(struct inode *);
2670 extern void __destroy_inode(struct inode *);
2671 extern struct inode *new_inode_pseudo(struct super_block *sb);
2672 extern struct inode *new_inode(struct super_block *sb);
2673 extern void free_inode_nonrcu(struct inode *inode);
2674 extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *);
2675 extern int file_remove_privs(struct file *);
2678 * This must be used for allocating filesystems specific inodes to set
2679 * up the inode reclaim context correctly.
2681 static inline void *
2682 alloc_inode_sb(struct super_block *sb, struct kmem_cache *cache, gfp_t gfp)
2684 return kmem_cache_alloc_lru(cache, &sb->s_inode_lru, gfp);
2687 extern void __insert_inode_hash(struct inode *, unsigned long hashval);
2688 static inline void insert_inode_hash(struct inode *inode)
2690 __insert_inode_hash(inode, inode->i_ino);
2693 extern void __remove_inode_hash(struct inode *);
2694 static inline void remove_inode_hash(struct inode *inode)
2696 if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash))
2697 __remove_inode_hash(inode);
2700 extern void inode_sb_list_add(struct inode *inode);
2701 extern void inode_add_lru(struct inode *inode);
2703 extern int sb_set_blocksize(struct super_block *, int);
2704 extern int sb_min_blocksize(struct super_block *, int);
2706 extern int generic_file_mmap(struct file *, struct vm_area_struct *);
2707 extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *);
2708 extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *);
2709 int generic_write_checks_count(struct kiocb *iocb, loff_t *count);
2710 extern int generic_write_check_limits(struct file *file, loff_t pos,
2712 extern int generic_file_rw_checks(struct file *file_in, struct file *file_out);
2713 ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *to,
2714 ssize_t already_read);
2715 extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *);
2716 extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *);
2717 extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *);
2718 extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *);
2719 ssize_t generic_perform_write(struct kiocb *, struct iov_iter *);
2721 ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos,
2723 ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos,
2725 ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb,
2726 struct iov_iter *iter);
2727 ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb,
2728 struct iov_iter *iter);
2731 ssize_t filemap_splice_read(struct file *in, loff_t *ppos,
2732 struct pipe_inode_info *pipe,
2733 size_t len, unsigned int flags);
2734 ssize_t direct_splice_read(struct file *in, loff_t *ppos,
2735 struct pipe_inode_info *pipe,
2736 size_t len, unsigned int flags);
2737 extern ssize_t generic_file_splice_read(struct file *, loff_t *,
2738 struct pipe_inode_info *, size_t, unsigned int);
2739 extern ssize_t iter_file_splice_write(struct pipe_inode_info *,
2740 struct file *, loff_t *, size_t, unsigned int);
2741 extern ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe,
2742 struct file *out, loff_t *, size_t len, unsigned int flags);
2743 extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
2744 loff_t *opos, size_t len, unsigned int flags);
2748 file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping);
2749 extern loff_t noop_llseek(struct file *file, loff_t offset, int whence);
2750 #define no_llseek NULL
2751 extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize);
2752 extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence);
2753 extern loff_t generic_file_llseek_size(struct file *file, loff_t offset,
2754 int whence, loff_t maxsize, loff_t eof);
2755 extern loff_t fixed_size_llseek(struct file *file, loff_t offset,
2756 int whence, loff_t size);
2757 extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t);
2758 extern loff_t no_seek_end_llseek(struct file *, loff_t, int);
2759 int rw_verify_area(int, struct file *, const loff_t *, size_t);
2760 extern int generic_file_open(struct inode * inode, struct file * filp);
2761 extern int nonseekable_open(struct inode * inode, struct file * filp);
2762 extern int stream_open(struct inode * inode, struct file * filp);
2765 typedef void (dio_submit_t)(struct bio *bio, struct inode *inode,
2766 loff_t file_offset);
2769 /* need locking between buffered and direct access */
2772 /* filesystem does not support filling holes */
2773 DIO_SKIP_HOLES = 0x02,
2776 ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
2777 struct block_device *bdev, struct iov_iter *iter,
2778 get_block_t get_block,
2779 dio_iodone_t end_io, dio_submit_t submit_io,
2782 static inline ssize_t blockdev_direct_IO(struct kiocb *iocb,
2783 struct inode *inode,
2784 struct iov_iter *iter,
2785 get_block_t get_block)
2787 return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter,
2788 get_block, NULL, NULL, DIO_LOCKING | DIO_SKIP_HOLES);
2792 void inode_dio_wait(struct inode *inode);
2795 * inode_dio_begin - signal start of a direct I/O requests
2796 * @inode: inode the direct I/O happens on
2798 * This is called once we've finished processing a direct I/O request,
2799 * and is used to wake up callers waiting for direct I/O to be quiesced.
2801 static inline void inode_dio_begin(struct inode *inode)
2803 atomic_inc(&inode->i_dio_count);
2807 * inode_dio_end - signal finish of a direct I/O requests
2808 * @inode: inode the direct I/O happens on
2810 * This is called once we've finished processing a direct I/O request,
2811 * and is used to wake up callers waiting for direct I/O to be quiesced.
2813 static inline void inode_dio_end(struct inode *inode)
2815 if (atomic_dec_and_test(&inode->i_dio_count))
2816 wake_up_bit(&inode->i_state, __I_DIO_WAKEUP);
2820 * Warn about a page cache invalidation failure diring a direct I/O write.
2822 void dio_warn_stale_pagecache(struct file *filp);
2824 extern void inode_set_flags(struct inode *inode, unsigned int flags,
2827 extern const struct file_operations generic_ro_fops;
2829 #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m))
2831 extern int readlink_copy(char __user *, int, const char *);
2832 extern int page_readlink(struct dentry *, char __user *, int);
2833 extern const char *page_get_link(struct dentry *, struct inode *,
2834 struct delayed_call *);
2835 extern void page_put_link(void *);
2836 extern int page_symlink(struct inode *inode, const char *symname, int len);
2837 extern const struct inode_operations page_symlink_inode_operations;
2838 extern void kfree_link(void *);
2839 void generic_fillattr(struct mnt_idmap *, struct inode *, struct kstat *);
2840 void generic_fill_statx_attr(struct inode *inode, struct kstat *stat);
2841 extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int);
2842 extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int);
2843 void __inode_add_bytes(struct inode *inode, loff_t bytes);
2844 void inode_add_bytes(struct inode *inode, loff_t bytes);
2845 void __inode_sub_bytes(struct inode *inode, loff_t bytes);
2846 void inode_sub_bytes(struct inode *inode, loff_t bytes);
2847 static inline loff_t __inode_get_bytes(struct inode *inode)
2849 return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes;
2851 loff_t inode_get_bytes(struct inode *inode);
2852 void inode_set_bytes(struct inode *inode, loff_t bytes);
2853 const char *simple_get_link(struct dentry *, struct inode *,
2854 struct delayed_call *);
2855 extern const struct inode_operations simple_symlink_inode_operations;
2857 extern int iterate_dir(struct file *, struct dir_context *);
2859 int vfs_fstatat(int dfd, const char __user *filename, struct kstat *stat,
2861 int vfs_fstat(int fd, struct kstat *stat);
2863 static inline int vfs_stat(const char __user *filename, struct kstat *stat)
2865 return vfs_fstatat(AT_FDCWD, filename, stat, 0);
2867 static inline int vfs_lstat(const char __user *name, struct kstat *stat)
2869 return vfs_fstatat(AT_FDCWD, name, stat, AT_SYMLINK_NOFOLLOW);
2872 extern const char *vfs_get_link(struct dentry *, struct delayed_call *);
2873 extern int vfs_readlink(struct dentry *, char __user *, int);
2875 extern struct file_system_type *get_filesystem(struct file_system_type *fs);
2876 extern void put_filesystem(struct file_system_type *fs);
2877 extern struct file_system_type *get_fs_type(const char *name);
2878 extern struct super_block *get_super(struct block_device *);
2879 extern struct super_block *get_active_super(struct block_device *bdev);
2880 extern void drop_super(struct super_block *sb);
2881 extern void drop_super_exclusive(struct super_block *sb);
2882 extern void iterate_supers(void (*)(struct super_block *, void *), void *);
2883 extern void iterate_supers_type(struct file_system_type *,
2884 void (*)(struct super_block *, void *), void *);
2886 extern int dcache_dir_open(struct inode *, struct file *);
2887 extern int dcache_dir_close(struct inode *, struct file *);
2888 extern loff_t dcache_dir_lseek(struct file *, loff_t, int);
2889 extern int dcache_readdir(struct file *, struct dir_context *);
2890 extern int simple_setattr(struct mnt_idmap *, struct dentry *,
2892 extern int simple_getattr(struct mnt_idmap *, const struct path *,
2893 struct kstat *, u32, unsigned int);
2894 extern int simple_statfs(struct dentry *, struct kstatfs *);
2895 extern int simple_open(struct inode *inode, struct file *file);
2896 extern int simple_link(struct dentry *, struct inode *, struct dentry *);
2897 extern int simple_unlink(struct inode *, struct dentry *);
2898 extern int simple_rmdir(struct inode *, struct dentry *);
2899 extern int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry,
2900 struct inode *new_dir, struct dentry *new_dentry);
2901 extern int simple_rename(struct mnt_idmap *, struct inode *,
2902 struct dentry *, struct inode *, struct dentry *,
2904 extern void simple_recursive_removal(struct dentry *,
2905 void (*callback)(struct dentry *));
2906 extern int noop_fsync(struct file *, loff_t, loff_t, int);
2907 extern ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter);
2908 extern int simple_empty(struct dentry *);
2909 extern int simple_write_begin(struct file *file, struct address_space *mapping,
2910 loff_t pos, unsigned len,
2911 struct page **pagep, void **fsdata);
2912 extern const struct address_space_operations ram_aops;
2913 extern int always_delete_dentry(const struct dentry *);
2914 extern struct inode *alloc_anon_inode(struct super_block *);
2915 extern int simple_nosetlease(struct file *, long, struct file_lock **, void **);
2916 extern const struct dentry_operations simple_dentry_operations;
2918 extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags);
2919 extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *);
2920 extern const struct file_operations simple_dir_operations;
2921 extern const struct inode_operations simple_dir_inode_operations;
2922 extern void make_empty_dir_inode(struct inode *inode);
2923 extern bool is_empty_dir_inode(struct inode *inode);
2924 struct tree_descr { const char *name; const struct file_operations *ops; int mode; };
2925 struct dentry *d_alloc_name(struct dentry *, const char *);
2926 extern int simple_fill_super(struct super_block *, unsigned long,
2927 const struct tree_descr *);
2928 extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count);
2929 extern void simple_release_fs(struct vfsmount **mount, int *count);
2931 extern ssize_t simple_read_from_buffer(void __user *to, size_t count,
2932 loff_t *ppos, const void *from, size_t available);
2933 extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
2934 const void __user *from, size_t count);
2936 extern int __generic_file_fsync(struct file *, loff_t, loff_t, int);
2937 extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
2939 extern int generic_check_addressable(unsigned, u64);
2941 extern void generic_set_encrypted_ci_d_ops(struct dentry *dentry);
2943 int may_setattr(struct mnt_idmap *idmap, struct inode *inode,
2944 unsigned int ia_valid);
2945 int setattr_prepare(struct mnt_idmap *, struct dentry *, struct iattr *);
2946 extern int inode_newsize_ok(const struct inode *, loff_t offset);
2947 void setattr_copy(struct mnt_idmap *, struct inode *inode,
2948 const struct iattr *attr);
2950 extern int file_update_time(struct file *file);
2952 static inline bool vma_is_dax(const struct vm_area_struct *vma)
2954 return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host);
2957 static inline bool vma_is_fsdax(struct vm_area_struct *vma)
2959 struct inode *inode;
2961 if (!IS_ENABLED(CONFIG_FS_DAX) || !vma->vm_file)
2963 if (!vma_is_dax(vma))
2965 inode = file_inode(vma->vm_file);
2966 if (S_ISCHR(inode->i_mode))
2967 return false; /* device-dax */
2971 static inline int iocb_flags(struct file *file)
2974 if (file->f_flags & O_APPEND)
2976 if (file->f_flags & O_DIRECT)
2978 if (file->f_flags & O_DSYNC)
2980 if (file->f_flags & __O_SYNC)
2985 static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags)
2987 int kiocb_flags = 0;
2989 /* make sure there's no overlap between RWF and private IOCB flags */
2990 BUILD_BUG_ON((__force int) RWF_SUPPORTED & IOCB_EVENTFD);
2994 if (unlikely(flags & ~RWF_SUPPORTED))
2997 if (flags & RWF_NOWAIT) {
2998 if (!(ki->ki_filp->f_mode & FMODE_NOWAIT))
3000 kiocb_flags |= IOCB_NOIO;
3002 kiocb_flags |= (__force int) (flags & RWF_SUPPORTED);
3003 if (flags & RWF_SYNC)
3004 kiocb_flags |= IOCB_DSYNC;
3006 ki->ki_flags |= kiocb_flags;
3010 static inline ino_t parent_ino(struct dentry *dentry)
3015 * Don't strictly need d_lock here? If the parent ino could change
3016 * then surely we'd have a deeper race in the caller?
3018 spin_lock(&dentry->d_lock);
3019 res = dentry->d_parent->d_inode->i_ino;
3020 spin_unlock(&dentry->d_lock);
3024 /* Transaction based IO helpers */
3027 * An argresp is stored in an allocated page and holds the
3028 * size of the argument or response, along with its content
3030 struct simple_transaction_argresp {
3035 #define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp))
3037 char *simple_transaction_get(struct file *file, const char __user *buf,
3039 ssize_t simple_transaction_read(struct file *file, char __user *buf,
3040 size_t size, loff_t *pos);
3041 int simple_transaction_release(struct inode *inode, struct file *file);
3043 void simple_transaction_set(struct file *file, size_t n);
3046 * simple attribute files
3048 * These attributes behave similar to those in sysfs:
3050 * Writing to an attribute immediately sets a value, an open file can be
3051 * written to multiple times.
3053 * Reading from an attribute creates a buffer from the value that might get
3054 * read with multiple read calls. When the attribute has been read
3055 * completely, no further read calls are possible until the file is opened
3058 * All attributes contain a text representation of a numeric value
3059 * that are accessed with the get() and set() functions.
3061 #define DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, __is_signed) \
3062 static int __fops ## _open(struct inode *inode, struct file *file) \
3064 __simple_attr_check_format(__fmt, 0ull); \
3065 return simple_attr_open(inode, file, __get, __set, __fmt); \
3067 static const struct file_operations __fops = { \
3068 .owner = THIS_MODULE, \
3069 .open = __fops ## _open, \
3070 .release = simple_attr_release, \
3071 .read = simple_attr_read, \
3072 .write = (__is_signed) ? simple_attr_write_signed : simple_attr_write, \
3073 .llseek = generic_file_llseek, \
3076 #define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt) \
3077 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, false)
3079 #define DEFINE_SIMPLE_ATTRIBUTE_SIGNED(__fops, __get, __set, __fmt) \
3080 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, true)
3082 static inline __printf(1, 2)
3083 void __simple_attr_check_format(const char *fmt, ...)
3085 /* don't do anything, just let the compiler check the arguments; */
3088 int simple_attr_open(struct inode *inode, struct file *file,
3089 int (*get)(void *, u64 *), int (*set)(void *, u64),
3091 int simple_attr_release(struct inode *inode, struct file *file);
3092 ssize_t simple_attr_read(struct file *file, char __user *buf,
3093 size_t len, loff_t *ppos);
3094 ssize_t simple_attr_write(struct file *file, const char __user *buf,
3095 size_t len, loff_t *ppos);
3096 ssize_t simple_attr_write_signed(struct file *file, const char __user *buf,
3097 size_t len, loff_t *ppos);
3100 int __init list_bdev_fs_names(char *buf, size_t size);
3102 #define __FMODE_EXEC ((__force int) FMODE_EXEC)
3103 #define __FMODE_NONOTIFY ((__force int) FMODE_NONOTIFY)
3105 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
3106 #define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \
3107 (flag & __FMODE_NONOTIFY)))
3109 static inline bool is_sxid(umode_t mode)
3111 return mode & (S_ISUID | S_ISGID);
3114 static inline int check_sticky(struct mnt_idmap *idmap,
3115 struct inode *dir, struct inode *inode)
3117 if (!(dir->i_mode & S_ISVTX))
3120 return __check_sticky(idmap, dir, inode);
3123 static inline void inode_has_no_xattr(struct inode *inode)
3125 if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC))
3126 inode->i_flags |= S_NOSEC;
3129 static inline bool is_root_inode(struct inode *inode)
3131 return inode == inode->i_sb->s_root->d_inode;
3134 static inline bool dir_emit(struct dir_context *ctx,
3135 const char *name, int namelen,
3136 u64 ino, unsigned type)
3138 return ctx->actor(ctx, name, namelen, ctx->pos, ino, type);
3140 static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx)
3142 return ctx->actor(ctx, ".", 1, ctx->pos,
3143 file->f_path.dentry->d_inode->i_ino, DT_DIR);
3145 static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx)
3147 return ctx->actor(ctx, "..", 2, ctx->pos,
3148 parent_ino(file->f_path.dentry), DT_DIR);
3150 static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx)
3152 if (ctx->pos == 0) {
3153 if (!dir_emit_dot(file, ctx))
3157 if (ctx->pos == 1) {
3158 if (!dir_emit_dotdot(file, ctx))
3164 static inline bool dir_relax(struct inode *inode)
3166 inode_unlock(inode);
3168 return !IS_DEADDIR(inode);
3171 static inline bool dir_relax_shared(struct inode *inode)
3173 inode_unlock_shared(inode);
3174 inode_lock_shared(inode);
3175 return !IS_DEADDIR(inode);
3178 extern bool path_noexec(const struct path *path);
3179 extern void inode_nohighmem(struct inode *inode);
3182 extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len,
3184 extern int generic_fadvise(struct file *file, loff_t offset, loff_t len,
3187 #endif /* _LINUX_FS_H */