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 /* The 'ki_filp' pointer is shared in a union for aio */
345 randomized_struct_fields_start
348 void (*ki_complete)(struct kiocb *iocb, long ret);
351 u16 ki_ioprio; /* See linux/ioprio.h */
352 struct wait_page_queue *ki_waitq; /* for async buffered IO */
353 randomized_struct_fields_end
356 static inline bool is_sync_kiocb(struct kiocb *kiocb)
358 return kiocb->ki_complete == NULL;
361 struct address_space_operations {
362 int (*writepage)(struct page *page, struct writeback_control *wbc);
363 int (*read_folio)(struct file *, struct folio *);
365 /* Write back some dirty pages from this mapping. */
366 int (*writepages)(struct address_space *, struct writeback_control *);
368 /* Mark a folio dirty. Return true if this dirtied it */
369 bool (*dirty_folio)(struct address_space *, struct folio *);
371 void (*readahead)(struct readahead_control *);
373 int (*write_begin)(struct file *, struct address_space *mapping,
374 loff_t pos, unsigned len,
375 struct page **pagep, void **fsdata);
376 int (*write_end)(struct file *, struct address_space *mapping,
377 loff_t pos, unsigned len, unsigned copied,
378 struct page *page, void *fsdata);
380 /* Unfortunately this kludge is needed for FIBMAP. Don't use it */
381 sector_t (*bmap)(struct address_space *, sector_t);
382 void (*invalidate_folio) (struct folio *, size_t offset, size_t len);
383 bool (*release_folio)(struct folio *, gfp_t);
384 void (*free_folio)(struct folio *folio);
385 ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter);
387 * migrate the contents of a folio to the specified target. If
388 * migrate_mode is MIGRATE_ASYNC, it must not block.
390 int (*migrate_folio)(struct address_space *, struct folio *dst,
391 struct folio *src, enum migrate_mode);
392 int (*launder_folio)(struct folio *);
393 bool (*is_partially_uptodate) (struct folio *, size_t from,
395 void (*is_dirty_writeback) (struct folio *, bool *dirty, bool *wb);
396 int (*error_remove_page)(struct address_space *, struct page *);
398 /* swapfile support */
399 int (*swap_activate)(struct swap_info_struct *sis, struct file *file,
401 void (*swap_deactivate)(struct file *file);
402 int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter);
405 extern const struct address_space_operations empty_aops;
408 * struct address_space - Contents of a cacheable, mappable object.
409 * @host: Owner, either the inode or the block_device.
410 * @i_pages: Cached pages.
411 * @invalidate_lock: Guards coherency between page cache contents and
412 * file offset->disk block mappings in the filesystem during invalidates.
413 * It is also used to block modification of page cache contents through
415 * @gfp_mask: Memory allocation flags to use for allocating pages.
416 * @i_mmap_writable: Number of VM_SHARED mappings.
417 * @nr_thps: Number of THPs in the pagecache (non-shmem only).
418 * @i_mmap: Tree of private and shared mappings.
419 * @i_mmap_rwsem: Protects @i_mmap and @i_mmap_writable.
420 * @nrpages: Number of page entries, protected by the i_pages lock.
421 * @writeback_index: Writeback starts here.
423 * @flags: Error bits and flags (AS_*).
424 * @wb_err: The most recent error which has occurred.
425 * @private_lock: For use by the owner of the address_space.
426 * @private_list: For use by the owner of the address_space.
427 * @private_data: For use by the owner of the address_space.
429 struct address_space {
431 struct xarray i_pages;
432 struct rw_semaphore invalidate_lock;
434 atomic_t i_mmap_writable;
435 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
436 /* number of thp, only for non-shmem files */
439 struct rb_root_cached i_mmap;
440 struct rw_semaphore i_mmap_rwsem;
441 unsigned long nrpages;
442 pgoff_t writeback_index;
443 const struct address_space_operations *a_ops;
446 spinlock_t private_lock;
447 struct list_head private_list;
449 } __attribute__((aligned(sizeof(long)))) __randomize_layout;
451 * On most architectures that alignment is already the case; but
452 * must be enforced here for CRIS, to let the least significant bit
453 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON.
456 /* XArray tags, for tagging dirty and writeback pages in the pagecache. */
457 #define PAGECACHE_TAG_DIRTY XA_MARK_0
458 #define PAGECACHE_TAG_WRITEBACK XA_MARK_1
459 #define PAGECACHE_TAG_TOWRITE XA_MARK_2
462 * Returns true if any of the pages in the mapping are marked with the tag.
464 static inline bool mapping_tagged(struct address_space *mapping, xa_mark_t tag)
466 return xa_marked(&mapping->i_pages, tag);
469 static inline void i_mmap_lock_write(struct address_space *mapping)
471 down_write(&mapping->i_mmap_rwsem);
474 static inline int i_mmap_trylock_write(struct address_space *mapping)
476 return down_write_trylock(&mapping->i_mmap_rwsem);
479 static inline void i_mmap_unlock_write(struct address_space *mapping)
481 up_write(&mapping->i_mmap_rwsem);
484 static inline int i_mmap_trylock_read(struct address_space *mapping)
486 return down_read_trylock(&mapping->i_mmap_rwsem);
489 static inline void i_mmap_lock_read(struct address_space *mapping)
491 down_read(&mapping->i_mmap_rwsem);
494 static inline void i_mmap_unlock_read(struct address_space *mapping)
496 up_read(&mapping->i_mmap_rwsem);
499 static inline void i_mmap_assert_locked(struct address_space *mapping)
501 lockdep_assert_held(&mapping->i_mmap_rwsem);
504 static inline void i_mmap_assert_write_locked(struct address_space *mapping)
506 lockdep_assert_held_write(&mapping->i_mmap_rwsem);
510 * Might pages of this file be mapped into userspace?
512 static inline int mapping_mapped(struct address_space *mapping)
514 return !RB_EMPTY_ROOT(&mapping->i_mmap.rb_root);
518 * Might pages of this file have been modified in userspace?
519 * Note that i_mmap_writable counts all VM_SHARED vmas: do_mmap
520 * marks vma as VM_SHARED if it is shared, and the file was opened for
521 * writing i.e. vma may be mprotected writable even if now readonly.
523 * If i_mmap_writable is negative, no new writable mappings are allowed. You
524 * can only deny writable mappings, if none exists right now.
526 static inline int mapping_writably_mapped(struct address_space *mapping)
528 return atomic_read(&mapping->i_mmap_writable) > 0;
531 static inline int mapping_map_writable(struct address_space *mapping)
533 return atomic_inc_unless_negative(&mapping->i_mmap_writable) ?
537 static inline void mapping_unmap_writable(struct address_space *mapping)
539 atomic_dec(&mapping->i_mmap_writable);
542 static inline int mapping_deny_writable(struct address_space *mapping)
544 return atomic_dec_unless_positive(&mapping->i_mmap_writable) ?
548 static inline void mapping_allow_writable(struct address_space *mapping)
550 atomic_inc(&mapping->i_mmap_writable);
554 * Use sequence counter to get consistent i_size on 32-bit processors.
556 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
557 #include <linux/seqlock.h>
558 #define __NEED_I_SIZE_ORDERED
559 #define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount)
561 #define i_size_ordered_init(inode) do { } while (0)
565 #define ACL_NOT_CACHED ((void *)(-1))
567 * ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to
568 * cache the ACL. This also means that ->get_acl() can be called in RCU mode
569 * with the LOOKUP_RCU flag.
571 #define ACL_DONT_CACHE ((void *)(-3))
573 static inline struct posix_acl *
574 uncached_acl_sentinel(struct task_struct *task)
576 return (void *)task + 1;
580 is_uncached_acl(struct posix_acl *acl)
582 return (long)acl & 1;
585 #define IOP_FASTPERM 0x0001
586 #define IOP_LOOKUP 0x0002
587 #define IOP_NOFOLLOW 0x0004
588 #define IOP_XATTR 0x0008
589 #define IOP_DEFAULT_READLINK 0x0010
591 struct fsnotify_mark_connector;
594 * Keep mostly read-only and often accessed (especially for
595 * the RCU path lookup and 'stat' data) fields at the beginning
596 * of the 'struct inode'
600 unsigned short i_opflags;
603 unsigned int i_flags;
605 #ifdef CONFIG_FS_POSIX_ACL
606 struct posix_acl *i_acl;
607 struct posix_acl *i_default_acl;
610 const struct inode_operations *i_op;
611 struct super_block *i_sb;
612 struct address_space *i_mapping;
614 #ifdef CONFIG_SECURITY
618 /* Stat data, not accessed from path walking */
621 * Filesystems may only read i_nlink directly. They shall use the
622 * following functions for modification:
624 * (set|clear|inc|drop)_nlink
625 * inode_(inc|dec)_link_count
628 const unsigned int i_nlink;
629 unsigned int __i_nlink;
633 struct timespec64 i_atime;
634 struct timespec64 i_mtime;
635 struct timespec64 i_ctime;
636 spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */
637 unsigned short i_bytes;
642 #ifdef __NEED_I_SIZE_ORDERED
643 seqcount_t i_size_seqcount;
647 unsigned long i_state;
648 struct rw_semaphore i_rwsem;
650 unsigned long dirtied_when; /* jiffies of first dirtying */
651 unsigned long dirtied_time_when;
653 struct hlist_node i_hash;
654 struct list_head i_io_list; /* backing dev IO list */
655 #ifdef CONFIG_CGROUP_WRITEBACK
656 struct bdi_writeback *i_wb; /* the associated cgroup wb */
658 /* foreign inode detection, see wbc_detach_inode() */
660 u16 i_wb_frn_avg_time;
661 u16 i_wb_frn_history;
663 struct list_head i_lru; /* inode LRU list */
664 struct list_head i_sb_list;
665 struct list_head i_wb_list; /* backing dev writeback list */
667 struct hlist_head i_dentry;
668 struct rcu_head i_rcu;
670 atomic64_t i_version;
671 atomic64_t i_sequence; /* see futex */
673 atomic_t i_dio_count;
674 atomic_t i_writecount;
675 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
676 atomic_t i_readcount; /* struct files open RO */
679 const struct file_operations *i_fop; /* former ->i_op->default_file_ops */
680 void (*free_inode)(struct inode *);
682 struct file_lock_context *i_flctx;
683 struct address_space i_data;
684 struct list_head i_devices;
686 struct pipe_inode_info *i_pipe;
694 #ifdef CONFIG_FSNOTIFY
695 __u32 i_fsnotify_mask; /* all events this inode cares about */
696 struct fsnotify_mark_connector __rcu *i_fsnotify_marks;
699 #ifdef CONFIG_FS_ENCRYPTION
700 struct fscrypt_info *i_crypt_info;
703 #ifdef CONFIG_FS_VERITY
704 struct fsverity_info *i_verity_info;
707 void *i_private; /* fs or device private pointer */
708 } __randomize_layout;
710 struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode);
712 static inline unsigned int i_blocksize(const struct inode *node)
714 return (1 << node->i_blkbits);
717 static inline int inode_unhashed(struct inode *inode)
719 return hlist_unhashed(&inode->i_hash);
723 * __mark_inode_dirty expects inodes to be hashed. Since we don't
724 * want special inodes in the fileset inode space, we make them
725 * appear hashed, but do not put on any lists. hlist_del()
726 * will work fine and require no locking.
728 static inline void inode_fake_hash(struct inode *inode)
730 hlist_add_fake(&inode->i_hash);
734 * inode->i_mutex nesting subclasses for the lock validator:
736 * 0: the object of the current VFS operation
740 * 4: second non-directory
741 * 5: second parent (when locking independent directories in rename)
743 * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two
744 * non-directories at once.
746 * The locking order between these classes is
747 * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory
749 enum inode_i_mutex_lock_class
759 static inline void inode_lock(struct inode *inode)
761 down_write(&inode->i_rwsem);
764 static inline void inode_unlock(struct inode *inode)
766 up_write(&inode->i_rwsem);
769 static inline void inode_lock_shared(struct inode *inode)
771 down_read(&inode->i_rwsem);
774 static inline void inode_unlock_shared(struct inode *inode)
776 up_read(&inode->i_rwsem);
779 static inline int inode_trylock(struct inode *inode)
781 return down_write_trylock(&inode->i_rwsem);
784 static inline int inode_trylock_shared(struct inode *inode)
786 return down_read_trylock(&inode->i_rwsem);
789 static inline int inode_is_locked(struct inode *inode)
791 return rwsem_is_locked(&inode->i_rwsem);
794 static inline void inode_lock_nested(struct inode *inode, unsigned subclass)
796 down_write_nested(&inode->i_rwsem, subclass);
799 static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass)
801 down_read_nested(&inode->i_rwsem, subclass);
804 static inline void filemap_invalidate_lock(struct address_space *mapping)
806 down_write(&mapping->invalidate_lock);
809 static inline void filemap_invalidate_unlock(struct address_space *mapping)
811 up_write(&mapping->invalidate_lock);
814 static inline void filemap_invalidate_lock_shared(struct address_space *mapping)
816 down_read(&mapping->invalidate_lock);
819 static inline int filemap_invalidate_trylock_shared(
820 struct address_space *mapping)
822 return down_read_trylock(&mapping->invalidate_lock);
825 static inline void filemap_invalidate_unlock_shared(
826 struct address_space *mapping)
828 up_read(&mapping->invalidate_lock);
831 void lock_two_nondirectories(struct inode *, struct inode*);
832 void unlock_two_nondirectories(struct inode *, struct inode*);
834 void filemap_invalidate_lock_two(struct address_space *mapping1,
835 struct address_space *mapping2);
836 void filemap_invalidate_unlock_two(struct address_space *mapping1,
837 struct address_space *mapping2);
841 * NOTE: in a 32bit arch with a preemptable kernel and
842 * an UP compile the i_size_read/write must be atomic
843 * with respect to the local cpu (unlike with preempt disabled),
844 * but they don't need to be atomic with respect to other cpus like in
845 * true SMP (so they need either to either locally disable irq around
846 * the read or for example on x86 they can be still implemented as a
847 * cmpxchg8b without the need of the lock prefix). For SMP compiles
848 * and 64bit archs it makes no difference if preempt is enabled or not.
850 static inline loff_t i_size_read(const struct inode *inode)
852 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
857 seq = read_seqcount_begin(&inode->i_size_seqcount);
858 i_size = inode->i_size;
859 } while (read_seqcount_retry(&inode->i_size_seqcount, seq));
861 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
865 i_size = inode->i_size;
869 return inode->i_size;
874 * NOTE: unlike i_size_read(), i_size_write() does need locking around it
875 * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount
876 * can be lost, resulting in subsequent i_size_read() calls spinning forever.
878 static inline void i_size_write(struct inode *inode, loff_t i_size)
880 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
882 write_seqcount_begin(&inode->i_size_seqcount);
883 inode->i_size = i_size;
884 write_seqcount_end(&inode->i_size_seqcount);
886 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
888 inode->i_size = i_size;
891 inode->i_size = i_size;
895 static inline unsigned iminor(const struct inode *inode)
897 return MINOR(inode->i_rdev);
900 static inline unsigned imajor(const struct inode *inode)
902 return MAJOR(inode->i_rdev);
906 rwlock_t lock; /* protects pid, uid, euid fields */
907 struct pid *pid; /* pid or -pgrp where SIGIO should be sent */
908 enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */
909 kuid_t uid, euid; /* uid/euid of process setting the owner */
910 int signum; /* posix.1b rt signal to be delivered on IO */
914 * struct file_ra_state - Track a file's readahead state.
915 * @start: Where the most recent readahead started.
916 * @size: Number of pages read in the most recent readahead.
917 * @async_size: Numer of pages that were/are not needed immediately
918 * and so were/are genuinely "ahead". Start next readahead when
919 * the first of these pages is accessed.
920 * @ra_pages: Maximum size of a readahead request, copied from the bdi.
921 * @mmap_miss: How many mmap accesses missed in the page cache.
922 * @prev_pos: The last byte in the most recent read request.
924 * When this structure is passed to ->readahead(), the "most recent"
925 * readahead means the current readahead.
927 struct file_ra_state {
930 unsigned int async_size;
931 unsigned int ra_pages;
932 unsigned int mmap_miss;
937 * Check if @index falls in the readahead windows.
939 static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index)
941 return (index >= ra->start &&
942 index < ra->start + ra->size);
947 struct llist_node f_llist;
948 struct rcu_head f_rcuhead;
949 unsigned int f_iocb_flags;
952 struct inode *f_inode; /* cached value */
953 const struct file_operations *f_op;
956 * Protects f_ep, f_flags.
957 * Must not be taken from IRQ context.
960 atomic_long_t f_count;
961 unsigned int f_flags;
963 struct mutex f_pos_lock;
965 struct fown_struct f_owner;
966 const struct cred *f_cred;
967 struct file_ra_state f_ra;
970 #ifdef CONFIG_SECURITY
973 /* needed for tty driver, and maybe others */
977 /* Used by fs/eventpoll.c to link all the hooks to this file */
978 struct hlist_head *f_ep;
979 #endif /* #ifdef CONFIG_EPOLL */
980 struct address_space *f_mapping;
982 errseq_t f_sb_err; /* for syncfs */
984 __attribute__((aligned(4))); /* lest something weird decides that 2 is OK */
989 /* file identifier */
990 unsigned char f_handle[];
993 static inline struct file *get_file(struct file *f)
995 atomic_long_inc(&f->f_count);
998 #define get_file_rcu(x) atomic_long_inc_not_zero(&(x)->f_count)
999 #define file_count(x) atomic_long_read(&(x)->f_count)
1001 #define MAX_NON_LFS ((1UL<<31) - 1)
1003 /* Page cache limit. The filesystems should put that into their s_maxbytes
1004 limits, otherwise bad things can happen in VM. */
1005 #if BITS_PER_LONG==32
1006 #define MAX_LFS_FILESIZE ((loff_t)ULONG_MAX << PAGE_SHIFT)
1007 #elif BITS_PER_LONG==64
1008 #define MAX_LFS_FILESIZE ((loff_t)LLONG_MAX)
1013 #define FL_DELEG 4 /* NFSv4 delegation */
1014 #define FL_ACCESS 8 /* not trying to lock, just looking */
1015 #define FL_EXISTS 16 /* when unlocking, test for existence */
1016 #define FL_LEASE 32 /* lease held on this file */
1017 #define FL_CLOSE 64 /* unlock on close */
1018 #define FL_SLEEP 128 /* A blocking lock */
1019 #define FL_DOWNGRADE_PENDING 256 /* Lease is being downgraded */
1020 #define FL_UNLOCK_PENDING 512 /* Lease is being broken */
1021 #define FL_OFDLCK 1024 /* lock is "owned" by struct file */
1022 #define FL_LAYOUT 2048 /* outstanding pNFS layout */
1023 #define FL_RECLAIM 4096 /* reclaiming from a reboot server */
1025 #define FL_CLOSE_POSIX (FL_POSIX | FL_CLOSE)
1028 * Special return value from posix_lock_file() and vfs_lock_file() for
1029 * asynchronous locking.
1031 #define FILE_LOCK_DEFERRED 1
1033 /* legacy typedef, should eventually be removed */
1034 typedef void *fl_owner_t;
1038 struct file_lock_operations {
1039 void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
1040 void (*fl_release_private)(struct file_lock *);
1043 struct lock_manager_operations {
1045 fl_owner_t (*lm_get_owner)(fl_owner_t);
1046 void (*lm_put_owner)(fl_owner_t);
1047 void (*lm_notify)(struct file_lock *); /* unblock callback */
1048 int (*lm_grant)(struct file_lock *, int);
1049 bool (*lm_break)(struct file_lock *);
1050 int (*lm_change)(struct file_lock *, int, struct list_head *);
1051 void (*lm_setup)(struct file_lock *, void **);
1052 bool (*lm_breaker_owns_lease)(struct file_lock *);
1053 bool (*lm_lock_expirable)(struct file_lock *cfl);
1054 void (*lm_expire_lock)(void);
1057 struct lock_manager {
1058 struct list_head list;
1060 * NFSv4 and up also want opens blocked during the grace period;
1067 void locks_start_grace(struct net *, struct lock_manager *);
1068 void locks_end_grace(struct lock_manager *);
1069 bool locks_in_grace(struct net *);
1070 bool opens_in_grace(struct net *);
1072 /* that will die - we need it for nfs_lock_info */
1073 #include <linux/nfs_fs_i.h>
1076 * struct file_lock represents a generic "file lock". It's used to represent
1077 * POSIX byte range locks, BSD (flock) locks, and leases. It's important to
1078 * note that the same struct is used to represent both a request for a lock and
1079 * the lock itself, but the same object is never used for both.
1081 * FIXME: should we create a separate "struct lock_request" to help distinguish
1084 * The varous i_flctx lists are ordered by:
1087 * 2) lock range start
1090 * Obviously, the last two criteria only matter for POSIX locks.
1093 struct file_lock *fl_blocker; /* The lock, that is blocking us */
1094 struct list_head fl_list; /* link into file_lock_context */
1095 struct hlist_node fl_link; /* node in global lists */
1096 struct list_head fl_blocked_requests; /* list of requests with
1097 * ->fl_blocker pointing here
1099 struct list_head fl_blocked_member; /* node in
1100 * ->fl_blocker->fl_blocked_requests
1102 fl_owner_t fl_owner;
1103 unsigned int fl_flags;
1104 unsigned char fl_type;
1105 unsigned int fl_pid;
1106 int fl_link_cpu; /* what cpu's list is this on? */
1107 wait_queue_head_t fl_wait;
1108 struct file *fl_file;
1112 struct fasync_struct * fl_fasync; /* for lease break notifications */
1113 /* for lease breaks: */
1114 unsigned long fl_break_time;
1115 unsigned long fl_downgrade_time;
1117 const struct file_lock_operations *fl_ops; /* Callbacks for filesystems */
1118 const struct lock_manager_operations *fl_lmops; /* Callbacks for lockmanagers */
1120 struct nfs_lock_info nfs_fl;
1121 struct nfs4_lock_info nfs4_fl;
1123 struct list_head link; /* link in AFS vnode's pending_locks list */
1124 int state; /* state of grant or error if -ve */
1125 unsigned int debug_id;
1128 } __randomize_layout;
1130 struct file_lock_context {
1131 spinlock_t flc_lock;
1132 struct list_head flc_flock;
1133 struct list_head flc_posix;
1134 struct list_head flc_lease;
1137 /* The following constant reflects the upper bound of the file/locking space */
1139 #define INT_LIMIT(x) (~((x)1 << (sizeof(x)*8 - 1)))
1140 #define OFFSET_MAX INT_LIMIT(loff_t)
1141 #define OFFT_OFFSET_MAX INT_LIMIT(off_t)
1144 extern void send_sigio(struct fown_struct *fown, int fd, int band);
1146 #define locks_inode(f) file_inode(f)
1148 #ifdef CONFIG_FILE_LOCKING
1149 extern int fcntl_getlk(struct file *, unsigned int, struct flock *);
1150 extern int fcntl_setlk(unsigned int, struct file *, unsigned int,
1153 #if BITS_PER_LONG == 32
1154 extern int fcntl_getlk64(struct file *, unsigned int, struct flock64 *);
1155 extern int fcntl_setlk64(unsigned int, struct file *, unsigned int,
1159 extern int fcntl_setlease(unsigned int fd, struct file *filp, long arg);
1160 extern int fcntl_getlease(struct file *filp);
1163 void locks_free_lock_context(struct inode *inode);
1164 void locks_free_lock(struct file_lock *fl);
1165 extern void locks_init_lock(struct file_lock *);
1166 extern struct file_lock * locks_alloc_lock(void);
1167 extern void locks_copy_lock(struct file_lock *, struct file_lock *);
1168 extern void locks_copy_conflock(struct file_lock *, struct file_lock *);
1169 extern void locks_remove_posix(struct file *, fl_owner_t);
1170 extern void locks_remove_file(struct file *);
1171 extern void locks_release_private(struct file_lock *);
1172 extern void posix_test_lock(struct file *, struct file_lock *);
1173 extern int posix_lock_file(struct file *, struct file_lock *, struct file_lock *);
1174 extern int locks_delete_block(struct file_lock *);
1175 extern int vfs_test_lock(struct file *, struct file_lock *);
1176 extern int vfs_lock_file(struct file *, unsigned int, struct file_lock *, struct file_lock *);
1177 extern int vfs_cancel_lock(struct file *filp, struct file_lock *fl);
1178 extern int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl);
1179 extern int __break_lease(struct inode *inode, unsigned int flags, unsigned int type);
1180 extern void lease_get_mtime(struct inode *, struct timespec64 *time);
1181 extern int generic_setlease(struct file *, long, struct file_lock **, void **priv);
1182 extern int vfs_setlease(struct file *, long, struct file_lock **, void **);
1183 extern int lease_modify(struct file_lock *, int, struct list_head *);
1185 struct notifier_block;
1186 extern int lease_register_notifier(struct notifier_block *);
1187 extern void lease_unregister_notifier(struct notifier_block *);
1189 struct files_struct;
1190 extern void show_fd_locks(struct seq_file *f,
1191 struct file *filp, struct files_struct *files);
1192 extern bool locks_owner_has_blockers(struct file_lock_context *flctx,
1194 #else /* !CONFIG_FILE_LOCKING */
1195 static inline int fcntl_getlk(struct file *file, unsigned int cmd,
1196 struct flock __user *user)
1201 static inline int fcntl_setlk(unsigned int fd, struct file *file,
1202 unsigned int cmd, struct flock __user *user)
1207 #if BITS_PER_LONG == 32
1208 static inline int fcntl_getlk64(struct file *file, unsigned int cmd,
1209 struct flock64 *user)
1214 static inline int fcntl_setlk64(unsigned int fd, struct file *file,
1215 unsigned int cmd, struct flock64 *user)
1220 static inline int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1225 static inline int fcntl_getlease(struct file *filp)
1231 locks_free_lock_context(struct inode *inode)
1235 static inline void locks_init_lock(struct file_lock *fl)
1240 static inline void locks_copy_conflock(struct file_lock *new, struct file_lock *fl)
1245 static inline void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
1250 static inline void locks_remove_posix(struct file *filp, fl_owner_t owner)
1255 static inline void locks_remove_file(struct file *filp)
1260 static inline void posix_test_lock(struct file *filp, struct file_lock *fl)
1265 static inline int posix_lock_file(struct file *filp, struct file_lock *fl,
1266 struct file_lock *conflock)
1271 static inline int locks_delete_block(struct file_lock *waiter)
1276 static inline int vfs_test_lock(struct file *filp, struct file_lock *fl)
1281 static inline int vfs_lock_file(struct file *filp, unsigned int cmd,
1282 struct file_lock *fl, struct file_lock *conf)
1287 static inline int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
1292 static inline int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1297 static inline int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
1302 static inline void lease_get_mtime(struct inode *inode,
1303 struct timespec64 *time)
1308 static inline int generic_setlease(struct file *filp, long arg,
1309 struct file_lock **flp, void **priv)
1314 static inline int vfs_setlease(struct file *filp, long arg,
1315 struct file_lock **lease, void **priv)
1320 static inline int lease_modify(struct file_lock *fl, int arg,
1321 struct list_head *dispose)
1326 struct files_struct;
1327 static inline void show_fd_locks(struct seq_file *f,
1328 struct file *filp, struct files_struct *files) {}
1329 static inline bool locks_owner_has_blockers(struct file_lock_context *flctx,
1334 #endif /* !CONFIG_FILE_LOCKING */
1336 static inline struct inode *file_inode(const struct file *f)
1341 static inline struct dentry *file_dentry(const struct file *file)
1343 return d_real(file->f_path.dentry, file_inode(file));
1346 static inline int locks_lock_file_wait(struct file *filp, struct file_lock *fl)
1348 return locks_lock_inode_wait(locks_inode(filp), fl);
1351 struct fasync_struct {
1355 struct fasync_struct *fa_next; /* singly linked list */
1356 struct file *fa_file;
1357 struct rcu_head fa_rcu;
1360 #define FASYNC_MAGIC 0x4601
1362 /* SMP safe fasync helpers: */
1363 extern int fasync_helper(int, struct file *, int, struct fasync_struct **);
1364 extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *);
1365 extern int fasync_remove_entry(struct file *, struct fasync_struct **);
1366 extern struct fasync_struct *fasync_alloc(void);
1367 extern void fasync_free(struct fasync_struct *);
1369 /* can be called from interrupts */
1370 extern void kill_fasync(struct fasync_struct **, int, int);
1372 extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force);
1373 extern int f_setown(struct file *filp, unsigned long arg, int force);
1374 extern void f_delown(struct file *filp);
1375 extern pid_t f_getown(struct file *filp);
1376 extern int send_sigurg(struct fown_struct *fown);
1379 * sb->s_flags. Note that these mirror the equivalent MS_* flags where
1380 * represented in both.
1382 #define SB_RDONLY 1 /* Mount read-only */
1383 #define SB_NOSUID 2 /* Ignore suid and sgid bits */
1384 #define SB_NODEV 4 /* Disallow access to device special files */
1385 #define SB_NOEXEC 8 /* Disallow program execution */
1386 #define SB_SYNCHRONOUS 16 /* Writes are synced at once */
1387 #define SB_MANDLOCK 64 /* Allow mandatory locks on an FS */
1388 #define SB_DIRSYNC 128 /* Directory modifications are synchronous */
1389 #define SB_NOATIME 1024 /* Do not update access times. */
1390 #define SB_NODIRATIME 2048 /* Do not update directory access times */
1391 #define SB_SILENT 32768
1392 #define SB_POSIXACL (1<<16) /* VFS does not apply the umask */
1393 #define SB_INLINECRYPT (1<<17) /* Use blk-crypto for encrypted files */
1394 #define SB_KERNMOUNT (1<<22) /* this is a kern_mount call */
1395 #define SB_I_VERSION (1<<23) /* Update inode I_version field */
1396 #define SB_LAZYTIME (1<<25) /* Update the on-disk [acm]times lazily */
1398 /* These sb flags are internal to the kernel */
1399 #define SB_SUBMOUNT (1<<26)
1400 #define SB_FORCE (1<<27)
1401 #define SB_NOSEC (1<<28)
1402 #define SB_BORN (1<<29)
1403 #define SB_ACTIVE (1<<30)
1404 #define SB_NOUSER (1<<31)
1406 /* These flags relate to encoding and casefolding */
1407 #define SB_ENC_STRICT_MODE_FL (1 << 0)
1409 #define sb_has_strict_encoding(sb) \
1410 (sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL)
1416 #define MNT_FORCE 0x00000001 /* Attempt to forcibily umount */
1417 #define MNT_DETACH 0x00000002 /* Just detach from the tree */
1418 #define MNT_EXPIRE 0x00000004 /* Mark for expiry */
1419 #define UMOUNT_NOFOLLOW 0x00000008 /* Don't follow symlink on umount */
1420 #define UMOUNT_UNUSED 0x80000000 /* Flag guaranteed to be unused */
1423 #define SB_I_CGROUPWB 0x00000001 /* cgroup-aware writeback enabled */
1424 #define SB_I_NOEXEC 0x00000002 /* Ignore executables on this fs */
1425 #define SB_I_NODEV 0x00000004 /* Ignore devices on this fs */
1426 #define SB_I_STABLE_WRITES 0x00000008 /* don't modify blks until WB is done */
1428 /* sb->s_iflags to limit user namespace mounts */
1429 #define SB_I_USERNS_VISIBLE 0x00000010 /* fstype already mounted */
1430 #define SB_I_IMA_UNVERIFIABLE_SIGNATURE 0x00000020
1431 #define SB_I_UNTRUSTED_MOUNTER 0x00000040
1433 #define SB_I_SKIP_SYNC 0x00000100 /* Skip superblock at global sync */
1434 #define SB_I_PERSB_BDI 0x00000200 /* has a per-sb bdi */
1435 #define SB_I_TS_EXPIRY_WARNED 0x00000400 /* warned about timestamp range expiry */
1436 #define SB_I_RETIRED 0x00000800 /* superblock shouldn't be reused */
1438 /* Possible states of 'frozen' field */
1440 SB_UNFROZEN = 0, /* FS is unfrozen */
1441 SB_FREEZE_WRITE = 1, /* Writes, dir ops, ioctls frozen */
1442 SB_FREEZE_PAGEFAULT = 2, /* Page faults stopped as well */
1443 SB_FREEZE_FS = 3, /* For internal FS use (e.g. to stop
1444 * internal threads if needed) */
1445 SB_FREEZE_COMPLETE = 4, /* ->freeze_fs finished successfully */
1448 #define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1)
1451 int frozen; /* Is sb frozen? */
1452 wait_queue_head_t wait_unfrozen; /* wait for thaw */
1453 struct percpu_rw_semaphore rw_sem[SB_FREEZE_LEVELS];
1456 struct super_block {
1457 struct list_head s_list; /* Keep this first */
1458 dev_t s_dev; /* search index; _not_ kdev_t */
1459 unsigned char s_blocksize_bits;
1460 unsigned long s_blocksize;
1461 loff_t s_maxbytes; /* Max file size */
1462 struct file_system_type *s_type;
1463 const struct super_operations *s_op;
1464 const struct dquot_operations *dq_op;
1465 const struct quotactl_ops *s_qcop;
1466 const struct export_operations *s_export_op;
1467 unsigned long s_flags;
1468 unsigned long s_iflags; /* internal SB_I_* flags */
1469 unsigned long s_magic;
1470 struct dentry *s_root;
1471 struct rw_semaphore s_umount;
1474 #ifdef CONFIG_SECURITY
1477 const struct xattr_handler **s_xattr;
1478 #ifdef CONFIG_FS_ENCRYPTION
1479 const struct fscrypt_operations *s_cop;
1480 struct key *s_master_keys; /* master crypto keys in use */
1482 #ifdef CONFIG_FS_VERITY
1483 const struct fsverity_operations *s_vop;
1485 #if IS_ENABLED(CONFIG_UNICODE)
1486 struct unicode_map *s_encoding;
1487 __u16 s_encoding_flags;
1489 struct hlist_bl_head s_roots; /* alternate root dentries for NFS */
1490 struct list_head s_mounts; /* list of mounts; _not_ for fs use */
1491 struct block_device *s_bdev;
1492 struct backing_dev_info *s_bdi;
1493 struct mtd_info *s_mtd;
1494 struct hlist_node s_instances;
1495 unsigned int s_quota_types; /* Bitmask of supported quota types */
1496 struct quota_info s_dquot; /* Diskquota specific options */
1498 struct sb_writers s_writers;
1501 * Keep s_fs_info, s_time_gran, s_fsnotify_mask, and
1502 * s_fsnotify_marks together for cache efficiency. They are frequently
1503 * accessed and rarely modified.
1505 void *s_fs_info; /* Filesystem private info */
1507 /* Granularity of c/m/atime in ns (cannot be worse than a second) */
1509 /* Time limits for c/m/atime in seconds */
1510 time64_t s_time_min;
1511 time64_t s_time_max;
1512 #ifdef CONFIG_FSNOTIFY
1513 __u32 s_fsnotify_mask;
1514 struct fsnotify_mark_connector __rcu *s_fsnotify_marks;
1517 char s_id[32]; /* Informational name */
1518 uuid_t s_uuid; /* UUID */
1520 unsigned int s_max_links;
1524 * The next field is for VFS *only*. No filesystems have any business
1525 * even looking at it. You had been warned.
1527 struct mutex s_vfs_rename_mutex; /* Kludge */
1530 * Filesystem subtype. If non-empty the filesystem type field
1531 * in /proc/mounts will be "type.subtype"
1533 const char *s_subtype;
1535 const struct dentry_operations *s_d_op; /* default d_op for dentries */
1537 struct shrinker s_shrink; /* per-sb shrinker handle */
1539 /* Number of inodes with nlink == 0 but still referenced */
1540 atomic_long_t s_remove_count;
1543 * Number of inode/mount/sb objects that are being watched, note that
1544 * inodes objects are currently double-accounted.
1546 atomic_long_t s_fsnotify_connectors;
1548 /* Being remounted read-only */
1549 int s_readonly_remount;
1551 /* per-sb errseq_t for reporting writeback errors via syncfs */
1554 /* AIO completions deferred from interrupt context */
1555 struct workqueue_struct *s_dio_done_wq;
1556 struct hlist_head s_pins;
1559 * Owning user namespace and default context in which to
1560 * interpret filesystem uids, gids, quotas, device nodes,
1561 * xattrs and security labels.
1563 struct user_namespace *s_user_ns;
1566 * The list_lru structure is essentially just a pointer to a table
1567 * of per-node lru lists, each of which has its own spinlock.
1568 * There is no need to put them into separate cachelines.
1570 struct list_lru s_dentry_lru;
1571 struct list_lru s_inode_lru;
1572 struct rcu_head rcu;
1573 struct work_struct destroy_work;
1575 struct mutex s_sync_lock; /* sync serialisation lock */
1578 * Indicates how deep in a filesystem stack this SB is
1582 /* s_inode_list_lock protects s_inodes */
1583 spinlock_t s_inode_list_lock ____cacheline_aligned_in_smp;
1584 struct list_head s_inodes; /* all inodes */
1586 spinlock_t s_inode_wblist_lock;
1587 struct list_head s_inodes_wb; /* writeback inodes */
1588 } __randomize_layout;
1590 static inline struct user_namespace *i_user_ns(const struct inode *inode)
1592 return inode->i_sb->s_user_ns;
1595 /* Helper functions so that in most cases filesystems will
1596 * not need to deal directly with kuid_t and kgid_t and can
1597 * instead deal with the raw numeric values that are stored
1598 * in the filesystem.
1600 static inline uid_t i_uid_read(const struct inode *inode)
1602 return from_kuid(i_user_ns(inode), inode->i_uid);
1605 static inline gid_t i_gid_read(const struct inode *inode)
1607 return from_kgid(i_user_ns(inode), inode->i_gid);
1610 static inline void i_uid_write(struct inode *inode, uid_t uid)
1612 inode->i_uid = make_kuid(i_user_ns(inode), uid);
1615 static inline void i_gid_write(struct inode *inode, gid_t gid)
1617 inode->i_gid = make_kgid(i_user_ns(inode), gid);
1621 * i_uid_into_mnt - map an inode's i_uid down into a mnt_userns
1622 * @mnt_userns: user namespace of the mount the inode was found from
1623 * @inode: inode to map
1625 * Note, this will eventually be removed completely in favor of the type-safe
1626 * i_uid_into_vfsuid().
1628 * Return: the inode's i_uid mapped down according to @mnt_userns.
1629 * If the inode's i_uid has no mapping INVALID_UID is returned.
1631 static inline kuid_t i_uid_into_mnt(struct user_namespace *mnt_userns,
1632 const struct inode *inode)
1634 return AS_KUIDT(make_vfsuid(mnt_userns, i_user_ns(inode), inode->i_uid));
1638 * i_uid_into_vfsuid - map an inode's i_uid down into a mnt_userns
1639 * @mnt_userns: user namespace of the mount the inode was found from
1640 * @inode: inode to map
1642 * Return: whe inode's i_uid mapped down according to @mnt_userns.
1643 * If the inode's i_uid has no mapping INVALID_VFSUID is returned.
1645 static inline vfsuid_t i_uid_into_vfsuid(struct user_namespace *mnt_userns,
1646 const struct inode *inode)
1648 return make_vfsuid(mnt_userns, i_user_ns(inode), inode->i_uid);
1652 * i_uid_needs_update - check whether inode's i_uid needs to be updated
1653 * @mnt_userns: user namespace of the mount the inode was found from
1654 * @attr: the new attributes of @inode
1655 * @inode: the inode to update
1657 * Check whether the $inode's i_uid field needs to be updated taking idmapped
1658 * mounts into account if the filesystem supports it.
1660 * Return: true if @inode's i_uid field needs to be updated, false if not.
1662 static inline bool i_uid_needs_update(struct user_namespace *mnt_userns,
1663 const struct iattr *attr,
1664 const struct inode *inode)
1666 return ((attr->ia_valid & ATTR_UID) &&
1667 !vfsuid_eq(attr->ia_vfsuid,
1668 i_uid_into_vfsuid(mnt_userns, inode)));
1672 * i_uid_update - update @inode's i_uid field
1673 * @mnt_userns: user namespace of the mount the inode was found from
1674 * @attr: the new attributes of @inode
1675 * @inode: the inode to update
1677 * Safely update @inode's i_uid field translating the vfsuid of any idmapped
1678 * mount into the filesystem kuid.
1680 static inline void i_uid_update(struct user_namespace *mnt_userns,
1681 const struct iattr *attr,
1682 struct inode *inode)
1684 if (attr->ia_valid & ATTR_UID)
1685 inode->i_uid = from_vfsuid(mnt_userns, i_user_ns(inode),
1690 * i_gid_into_mnt - map an inode's i_gid down into a mnt_userns
1691 * @mnt_userns: user namespace of the mount the inode was found from
1692 * @inode: inode to map
1694 * Note, this will eventually be removed completely in favor of the type-safe
1695 * i_gid_into_vfsgid().
1697 * Return: the inode's i_gid mapped down according to @mnt_userns.
1698 * If the inode's i_gid has no mapping INVALID_GID is returned.
1700 static inline kgid_t i_gid_into_mnt(struct user_namespace *mnt_userns,
1701 const struct inode *inode)
1703 return AS_KGIDT(make_vfsgid(mnt_userns, i_user_ns(inode), inode->i_gid));
1707 * i_gid_into_vfsgid - map an inode's i_gid down into a mnt_userns
1708 * @mnt_userns: user namespace of the mount the inode was found from
1709 * @inode: inode to map
1711 * Return: the inode's i_gid mapped down according to @mnt_userns.
1712 * If the inode's i_gid has no mapping INVALID_VFSGID is returned.
1714 static inline vfsgid_t i_gid_into_vfsgid(struct user_namespace *mnt_userns,
1715 const struct inode *inode)
1717 return make_vfsgid(mnt_userns, i_user_ns(inode), inode->i_gid);
1721 * i_gid_needs_update - check whether inode's i_gid needs to be updated
1722 * @mnt_userns: user namespace of the mount the inode was found from
1723 * @attr: the new attributes of @inode
1724 * @inode: the inode to update
1726 * Check whether the $inode's i_gid field needs to be updated taking idmapped
1727 * mounts into account if the filesystem supports it.
1729 * Return: true if @inode's i_gid field needs to be updated, false if not.
1731 static inline bool i_gid_needs_update(struct user_namespace *mnt_userns,
1732 const struct iattr *attr,
1733 const struct inode *inode)
1735 return ((attr->ia_valid & ATTR_GID) &&
1736 !vfsgid_eq(attr->ia_vfsgid,
1737 i_gid_into_vfsgid(mnt_userns, inode)));
1741 * i_gid_update - update @inode's i_gid field
1742 * @mnt_userns: user namespace of the mount the inode was found from
1743 * @attr: the new attributes of @inode
1744 * @inode: the inode to update
1746 * Safely update @inode's i_gid field translating the vfsgid of any idmapped
1747 * mount into the filesystem kgid.
1749 static inline void i_gid_update(struct user_namespace *mnt_userns,
1750 const struct iattr *attr,
1751 struct inode *inode)
1753 if (attr->ia_valid & ATTR_GID)
1754 inode->i_gid = from_vfsgid(mnt_userns, i_user_ns(inode),
1759 * inode_fsuid_set - initialize inode's i_uid field with callers fsuid
1760 * @inode: inode to initialize
1761 * @mnt_userns: user namespace of the mount the inode was found from
1763 * Initialize the i_uid field of @inode. If the inode was found/created via
1764 * an idmapped mount map the caller's fsuid according to @mnt_users.
1766 static inline void inode_fsuid_set(struct inode *inode,
1767 struct user_namespace *mnt_userns)
1769 inode->i_uid = mapped_fsuid(mnt_userns, i_user_ns(inode));
1773 * inode_fsgid_set - initialize inode's i_gid field with callers fsgid
1774 * @inode: inode to initialize
1775 * @mnt_userns: user namespace of the mount the inode was found from
1777 * Initialize the i_gid field of @inode. If the inode was found/created via
1778 * an idmapped mount map the caller's fsgid according to @mnt_users.
1780 static inline void inode_fsgid_set(struct inode *inode,
1781 struct user_namespace *mnt_userns)
1783 inode->i_gid = mapped_fsgid(mnt_userns, i_user_ns(inode));
1787 * fsuidgid_has_mapping() - check whether caller's fsuid/fsgid is mapped
1788 * @sb: the superblock we want a mapping in
1789 * @mnt_userns: user namespace of the relevant mount
1791 * Check whether the caller's fsuid and fsgid have a valid mapping in the
1792 * s_user_ns of the superblock @sb. If the caller is on an idmapped mount map
1793 * the caller's fsuid and fsgid according to the @mnt_userns first.
1795 * Return: true if fsuid and fsgid is mapped, false if not.
1797 static inline bool fsuidgid_has_mapping(struct super_block *sb,
1798 struct user_namespace *mnt_userns)
1800 struct user_namespace *fs_userns = sb->s_user_ns;
1804 kuid = mapped_fsuid(mnt_userns, fs_userns);
1805 if (!uid_valid(kuid))
1807 kgid = mapped_fsgid(mnt_userns, fs_userns);
1808 if (!gid_valid(kgid))
1810 return kuid_has_mapping(fs_userns, kuid) &&
1811 kgid_has_mapping(fs_userns, kgid);
1814 extern struct timespec64 current_time(struct inode *inode);
1817 * Snapshotting support.
1821 * These are internal functions, please use sb_start_{write,pagefault,intwrite}
1824 static inline void __sb_end_write(struct super_block *sb, int level)
1826 percpu_up_read(sb->s_writers.rw_sem + level-1);
1829 static inline void __sb_start_write(struct super_block *sb, int level)
1831 percpu_down_read(sb->s_writers.rw_sem + level - 1);
1834 static inline bool __sb_start_write_trylock(struct super_block *sb, int level)
1836 return percpu_down_read_trylock(sb->s_writers.rw_sem + level - 1);
1839 #define __sb_writers_acquired(sb, lev) \
1840 percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1841 #define __sb_writers_release(sb, lev) \
1842 percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1844 static inline bool sb_write_started(const struct super_block *sb)
1846 return lockdep_is_held_type(sb->s_writers.rw_sem + SB_FREEZE_WRITE - 1, 1);
1850 * sb_end_write - drop write access to a superblock
1851 * @sb: the super we wrote to
1853 * Decrement number of writers to the filesystem. Wake up possible waiters
1854 * wanting to freeze the filesystem.
1856 static inline void sb_end_write(struct super_block *sb)
1858 __sb_end_write(sb, SB_FREEZE_WRITE);
1862 * sb_end_pagefault - drop write access to a superblock from a page fault
1863 * @sb: the super we wrote to
1865 * Decrement number of processes handling write page fault to the filesystem.
1866 * Wake up possible waiters wanting to freeze the filesystem.
1868 static inline void sb_end_pagefault(struct super_block *sb)
1870 __sb_end_write(sb, SB_FREEZE_PAGEFAULT);
1874 * sb_end_intwrite - drop write access to a superblock for internal fs purposes
1875 * @sb: the super we wrote to
1877 * Decrement fs-internal number of writers to the filesystem. Wake up possible
1878 * waiters wanting to freeze the filesystem.
1880 static inline void sb_end_intwrite(struct super_block *sb)
1882 __sb_end_write(sb, SB_FREEZE_FS);
1886 * sb_start_write - get write access to a superblock
1887 * @sb: the super we write to
1889 * When a process wants to write data or metadata to a file system (i.e. dirty
1890 * a page or an inode), it should embed the operation in a sb_start_write() -
1891 * sb_end_write() pair to get exclusion against file system freezing. This
1892 * function increments number of writers preventing freezing. If the file
1893 * system is already frozen, the function waits until the file system is
1896 * Since freeze protection behaves as a lock, users have to preserve
1897 * ordering of freeze protection and other filesystem locks. Generally,
1898 * freeze protection should be the outermost lock. In particular, we have:
1901 * -> i_mutex (write path, truncate, directory ops, ...)
1902 * -> s_umount (freeze_super, thaw_super)
1904 static inline void sb_start_write(struct super_block *sb)
1906 __sb_start_write(sb, SB_FREEZE_WRITE);
1909 static inline bool sb_start_write_trylock(struct super_block *sb)
1911 return __sb_start_write_trylock(sb, SB_FREEZE_WRITE);
1915 * sb_start_pagefault - get write access to a superblock from a page fault
1916 * @sb: the super we write to
1918 * When a process starts handling write page fault, it should embed the
1919 * operation into sb_start_pagefault() - sb_end_pagefault() pair to get
1920 * exclusion against file system freezing. This is needed since the page fault
1921 * is going to dirty a page. This function increments number of running page
1922 * faults preventing freezing. If the file system is already frozen, the
1923 * function waits until the file system is thawed.
1925 * Since page fault freeze protection behaves as a lock, users have to preserve
1926 * ordering of freeze protection and other filesystem locks. It is advised to
1927 * put sb_start_pagefault() close to mmap_lock in lock ordering. Page fault
1928 * handling code implies lock dependency:
1931 * -> sb_start_pagefault
1933 static inline void sb_start_pagefault(struct super_block *sb)
1935 __sb_start_write(sb, SB_FREEZE_PAGEFAULT);
1939 * sb_start_intwrite - get write access to a superblock for internal fs purposes
1940 * @sb: the super we write to
1942 * This is the third level of protection against filesystem freezing. It is
1943 * free for use by a filesystem. The only requirement is that it must rank
1944 * below sb_start_pagefault.
1946 * For example filesystem can call sb_start_intwrite() when starting a
1947 * transaction which somewhat eases handling of freezing for internal sources
1948 * of filesystem changes (internal fs threads, discarding preallocation on file
1951 static inline void sb_start_intwrite(struct super_block *sb)
1953 __sb_start_write(sb, SB_FREEZE_FS);
1956 static inline bool sb_start_intwrite_trylock(struct super_block *sb)
1958 return __sb_start_write_trylock(sb, SB_FREEZE_FS);
1961 bool inode_owner_or_capable(struct user_namespace *mnt_userns,
1962 const struct inode *inode);
1965 * VFS helper functions..
1967 int vfs_create(struct user_namespace *, struct inode *,
1968 struct dentry *, umode_t, bool);
1969 int vfs_mkdir(struct user_namespace *, struct inode *,
1970 struct dentry *, umode_t);
1971 int vfs_mknod(struct user_namespace *, struct inode *, struct dentry *,
1973 int vfs_symlink(struct user_namespace *, struct inode *,
1974 struct dentry *, const char *);
1975 int vfs_link(struct dentry *, struct user_namespace *, struct inode *,
1976 struct dentry *, struct inode **);
1977 int vfs_rmdir(struct user_namespace *, struct inode *, struct dentry *);
1978 int vfs_unlink(struct user_namespace *, struct inode *, struct dentry *,
1982 * struct renamedata - contains all information required for renaming
1983 * @old_mnt_userns: old user namespace of the mount the inode was found from
1984 * @old_dir: parent of source
1985 * @old_dentry: source
1986 * @new_mnt_userns: new user namespace of the mount the inode was found from
1987 * @new_dir: parent of destination
1988 * @new_dentry: destination
1989 * @delegated_inode: returns an inode needing a delegation break
1990 * @flags: rename flags
1993 struct user_namespace *old_mnt_userns;
1994 struct inode *old_dir;
1995 struct dentry *old_dentry;
1996 struct user_namespace *new_mnt_userns;
1997 struct inode *new_dir;
1998 struct dentry *new_dentry;
1999 struct inode **delegated_inode;
2001 } __randomize_layout;
2003 int vfs_rename(struct renamedata *);
2005 static inline int vfs_whiteout(struct user_namespace *mnt_userns,
2006 struct inode *dir, struct dentry *dentry)
2008 return vfs_mknod(mnt_userns, dir, dentry, S_IFCHR | WHITEOUT_MODE,
2012 struct dentry *vfs_tmpfile(struct user_namespace *mnt_userns,
2013 struct dentry *dentry, umode_t mode, int open_flag);
2015 int vfs_mkobj(struct dentry *, umode_t,
2016 int (*f)(struct dentry *, umode_t, void *),
2019 int vfs_fchown(struct file *file, uid_t user, gid_t group);
2020 int vfs_fchmod(struct file *file, umode_t mode);
2021 int vfs_utimes(const struct path *path, struct timespec64 *times);
2023 extern long vfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
2025 #ifdef CONFIG_COMPAT
2026 extern long compat_ptr_ioctl(struct file *file, unsigned int cmd,
2029 #define compat_ptr_ioctl NULL
2033 * VFS file helper functions.
2035 void inode_init_owner(struct user_namespace *mnt_userns, struct inode *inode,
2036 const struct inode *dir, umode_t mode);
2037 extern bool may_open_dev(const struct path *path);
2040 * This is the "filldir" function type, used by readdir() to let
2041 * the kernel specify what kind of dirent layout it wants to have.
2042 * This allows the kernel to read directories into kernel space or
2043 * to have different dirent layouts depending on the binary type.
2046 typedef int (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64,
2049 struct dir_context {
2055 * These flags let !MMU mmap() govern direct device mapping vs immediate
2056 * copying more easily for MAP_PRIVATE, especially for ROM filesystems.
2058 * NOMMU_MAP_COPY: Copy can be mapped (MAP_PRIVATE)
2059 * NOMMU_MAP_DIRECT: Can be mapped directly (MAP_SHARED)
2060 * NOMMU_MAP_READ: Can be mapped for reading
2061 * NOMMU_MAP_WRITE: Can be mapped for writing
2062 * NOMMU_MAP_EXEC: Can be mapped for execution
2064 #define NOMMU_MAP_COPY 0x00000001
2065 #define NOMMU_MAP_DIRECT 0x00000008
2066 #define NOMMU_MAP_READ VM_MAYREAD
2067 #define NOMMU_MAP_WRITE VM_MAYWRITE
2068 #define NOMMU_MAP_EXEC VM_MAYEXEC
2070 #define NOMMU_VMFLAGS \
2071 (NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC)
2074 * These flags control the behavior of the remap_file_range function pointer.
2075 * If it is called with len == 0 that means "remap to end of source file".
2076 * See Documentation/filesystems/vfs.rst for more details about this call.
2078 * REMAP_FILE_DEDUP: only remap if contents identical (i.e. deduplicate)
2079 * REMAP_FILE_CAN_SHORTEN: caller can handle a shortened request
2081 #define REMAP_FILE_DEDUP (1 << 0)
2082 #define REMAP_FILE_CAN_SHORTEN (1 << 1)
2085 * These flags signal that the caller is ok with altering various aspects of
2086 * the behavior of the remap operation. The changes must be made by the
2087 * implementation; the vfs remap helper functions can take advantage of them.
2088 * Flags in this category exist to preserve the quirky behavior of the hoisted
2089 * btrfs clone/dedupe ioctls.
2091 #define REMAP_FILE_ADVISORY (REMAP_FILE_CAN_SHORTEN)
2094 struct io_uring_cmd;
2096 struct file_operations {
2097 struct module *owner;
2098 loff_t (*llseek) (struct file *, loff_t, int);
2099 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
2100 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
2101 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
2102 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
2103 int (*iopoll)(struct kiocb *kiocb, struct io_comp_batch *,
2104 unsigned int flags);
2105 int (*iterate) (struct file *, struct dir_context *);
2106 int (*iterate_shared) (struct file *, struct dir_context *);
2107 __poll_t (*poll) (struct file *, struct poll_table_struct *);
2108 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
2109 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
2110 int (*mmap) (struct file *, struct vm_area_struct *);
2111 unsigned long mmap_supported_flags;
2112 int (*open) (struct inode *, struct file *);
2113 int (*flush) (struct file *, fl_owner_t id);
2114 int (*release) (struct inode *, struct file *);
2115 int (*fsync) (struct file *, loff_t, loff_t, int datasync);
2116 int (*fasync) (int, struct file *, int);
2117 int (*lock) (struct file *, int, struct file_lock *);
2118 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
2119 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2120 int (*check_flags)(int);
2121 int (*flock) (struct file *, int, struct file_lock *);
2122 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
2123 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
2124 int (*setlease)(struct file *, long, struct file_lock **, void **);
2125 long (*fallocate)(struct file *file, int mode, loff_t offset,
2127 void (*show_fdinfo)(struct seq_file *m, struct file *f);
2129 unsigned (*mmap_capabilities)(struct file *);
2131 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *,
2132 loff_t, size_t, unsigned int);
2133 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in,
2134 struct file *file_out, loff_t pos_out,
2135 loff_t len, unsigned int remap_flags);
2136 int (*fadvise)(struct file *, loff_t, loff_t, int);
2137 int (*uring_cmd)(struct io_uring_cmd *ioucmd, unsigned int issue_flags);
2138 } __randomize_layout;
2140 struct inode_operations {
2141 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
2142 const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *);
2143 int (*permission) (struct user_namespace *, struct inode *, int);
2144 struct posix_acl * (*get_acl)(struct inode *, int, bool);
2146 int (*readlink) (struct dentry *, char __user *,int);
2148 int (*create) (struct user_namespace *, struct inode *,struct dentry *,
2150 int (*link) (struct dentry *,struct inode *,struct dentry *);
2151 int (*unlink) (struct inode *,struct dentry *);
2152 int (*symlink) (struct user_namespace *, struct inode *,struct dentry *,
2154 int (*mkdir) (struct user_namespace *, struct inode *,struct dentry *,
2156 int (*rmdir) (struct inode *,struct dentry *);
2157 int (*mknod) (struct user_namespace *, struct inode *,struct dentry *,
2159 int (*rename) (struct user_namespace *, struct inode *, struct dentry *,
2160 struct inode *, struct dentry *, unsigned int);
2161 int (*setattr) (struct user_namespace *, struct dentry *,
2163 int (*getattr) (struct user_namespace *, const struct path *,
2164 struct kstat *, u32, unsigned int);
2165 ssize_t (*listxattr) (struct dentry *, char *, size_t);
2166 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start,
2168 int (*update_time)(struct inode *, struct timespec64 *, int);
2169 int (*atomic_open)(struct inode *, struct dentry *,
2170 struct file *, unsigned open_flag,
2171 umode_t create_mode);
2172 int (*tmpfile) (struct user_namespace *, struct inode *,
2173 struct dentry *, umode_t);
2174 int (*set_acl)(struct user_namespace *, struct inode *,
2175 struct posix_acl *, int);
2176 int (*fileattr_set)(struct user_namespace *mnt_userns,
2177 struct dentry *dentry, struct fileattr *fa);
2178 int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa);
2179 } ____cacheline_aligned;
2181 static inline ssize_t call_read_iter(struct file *file, struct kiocb *kio,
2182 struct iov_iter *iter)
2184 return file->f_op->read_iter(kio, iter);
2187 static inline ssize_t call_write_iter(struct file *file, struct kiocb *kio,
2188 struct iov_iter *iter)
2190 return file->f_op->write_iter(kio, iter);
2193 static inline int call_mmap(struct file *file, struct vm_area_struct *vma)
2195 return file->f_op->mmap(file, vma);
2198 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
2199 extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *);
2200 extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *,
2201 loff_t, size_t, unsigned int);
2202 extern ssize_t generic_copy_file_range(struct file *file_in, loff_t pos_in,
2203 struct file *file_out, loff_t pos_out,
2204 size_t len, unsigned int flags);
2205 int __generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
2206 struct file *file_out, loff_t pos_out,
2207 loff_t *len, unsigned int remap_flags,
2208 const struct iomap_ops *dax_read_ops);
2209 int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
2210 struct file *file_out, loff_t pos_out,
2211 loff_t *count, unsigned int remap_flags);
2212 extern loff_t do_clone_file_range(struct file *file_in, loff_t pos_in,
2213 struct file *file_out, loff_t pos_out,
2214 loff_t len, unsigned int remap_flags);
2215 extern loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in,
2216 struct file *file_out, loff_t pos_out,
2217 loff_t len, unsigned int remap_flags);
2218 extern int vfs_dedupe_file_range(struct file *file,
2219 struct file_dedupe_range *same);
2220 extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos,
2221 struct file *dst_file, loff_t dst_pos,
2222 loff_t len, unsigned int remap_flags);
2225 struct super_operations {
2226 struct inode *(*alloc_inode)(struct super_block *sb);
2227 void (*destroy_inode)(struct inode *);
2228 void (*free_inode)(struct inode *);
2230 void (*dirty_inode) (struct inode *, int flags);
2231 int (*write_inode) (struct inode *, struct writeback_control *wbc);
2232 int (*drop_inode) (struct inode *);
2233 void (*evict_inode) (struct inode *);
2234 void (*put_super) (struct super_block *);
2235 int (*sync_fs)(struct super_block *sb, int wait);
2236 int (*freeze_super) (struct super_block *);
2237 int (*freeze_fs) (struct super_block *);
2238 int (*thaw_super) (struct super_block *);
2239 int (*unfreeze_fs) (struct super_block *);
2240 int (*statfs) (struct dentry *, struct kstatfs *);
2241 int (*remount_fs) (struct super_block *, int *, char *);
2242 void (*umount_begin) (struct super_block *);
2244 int (*show_options)(struct seq_file *, struct dentry *);
2245 int (*show_devname)(struct seq_file *, struct dentry *);
2246 int (*show_path)(struct seq_file *, struct dentry *);
2247 int (*show_stats)(struct seq_file *, struct dentry *);
2249 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
2250 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
2251 struct dquot **(*get_dquots)(struct inode *);
2253 long (*nr_cached_objects)(struct super_block *,
2254 struct shrink_control *);
2255 long (*free_cached_objects)(struct super_block *,
2256 struct shrink_control *);
2260 * Inode flags - they have no relation to superblock flags now
2262 #define S_SYNC (1 << 0) /* Writes are synced at once */
2263 #define S_NOATIME (1 << 1) /* Do not update access times */
2264 #define S_APPEND (1 << 2) /* Append-only file */
2265 #define S_IMMUTABLE (1 << 3) /* Immutable file */
2266 #define S_DEAD (1 << 4) /* removed, but still open directory */
2267 #define S_NOQUOTA (1 << 5) /* Inode is not counted to quota */
2268 #define S_DIRSYNC (1 << 6) /* Directory modifications are synchronous */
2269 #define S_NOCMTIME (1 << 7) /* Do not update file c/mtime */
2270 #define S_SWAPFILE (1 << 8) /* Do not truncate: swapon got its bmaps */
2271 #define S_PRIVATE (1 << 9) /* Inode is fs-internal */
2272 #define S_IMA (1 << 10) /* Inode has an associated IMA struct */
2273 #define S_AUTOMOUNT (1 << 11) /* Automount/referral quasi-directory */
2274 #define S_NOSEC (1 << 12) /* no suid or xattr security attributes */
2275 #ifdef CONFIG_FS_DAX
2276 #define S_DAX (1 << 13) /* Direct Access, avoiding the page cache */
2278 #define S_DAX 0 /* Make all the DAX code disappear */
2280 #define S_ENCRYPTED (1 << 14) /* Encrypted file (using fs/crypto/) */
2281 #define S_CASEFOLD (1 << 15) /* Casefolded file */
2282 #define S_VERITY (1 << 16) /* Verity file (using fs/verity/) */
2283 #define S_KERNEL_FILE (1 << 17) /* File is in use by the kernel (eg. fs/cachefiles) */
2286 * Note that nosuid etc flags are inode-specific: setting some file-system
2287 * flags just means all the inodes inherit those flags by default. It might be
2288 * possible to override it selectively if you really wanted to with some
2289 * ioctl() that is not currently implemented.
2291 * Exception: SB_RDONLY is always applied to the entire file system.
2293 * Unfortunately, it is possible to change a filesystems flags with it mounted
2294 * with files in use. This means that all of the inodes will not have their
2295 * i_flags updated. Hence, i_flags no longer inherit the superblock mount
2296 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org
2298 #define __IS_FLG(inode, flg) ((inode)->i_sb->s_flags & (flg))
2300 static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; }
2301 #define IS_RDONLY(inode) sb_rdonly((inode)->i_sb)
2302 #define IS_SYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS) || \
2303 ((inode)->i_flags & S_SYNC))
2304 #define IS_DIRSYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS|SB_DIRSYNC) || \
2305 ((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
2306 #define IS_MANDLOCK(inode) __IS_FLG(inode, SB_MANDLOCK)
2307 #define IS_NOATIME(inode) __IS_FLG(inode, SB_RDONLY|SB_NOATIME)
2308 #define IS_I_VERSION(inode) __IS_FLG(inode, SB_I_VERSION)
2310 #define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA)
2311 #define IS_APPEND(inode) ((inode)->i_flags & S_APPEND)
2312 #define IS_IMMUTABLE(inode) ((inode)->i_flags & S_IMMUTABLE)
2313 #define IS_POSIXACL(inode) __IS_FLG(inode, SB_POSIXACL)
2315 #define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD)
2316 #define IS_NOCMTIME(inode) ((inode)->i_flags & S_NOCMTIME)
2317 #define IS_SWAPFILE(inode) ((inode)->i_flags & S_SWAPFILE)
2318 #define IS_PRIVATE(inode) ((inode)->i_flags & S_PRIVATE)
2319 #define IS_IMA(inode) ((inode)->i_flags & S_IMA)
2320 #define IS_AUTOMOUNT(inode) ((inode)->i_flags & S_AUTOMOUNT)
2321 #define IS_NOSEC(inode) ((inode)->i_flags & S_NOSEC)
2322 #define IS_DAX(inode) ((inode)->i_flags & S_DAX)
2323 #define IS_ENCRYPTED(inode) ((inode)->i_flags & S_ENCRYPTED)
2324 #define IS_CASEFOLDED(inode) ((inode)->i_flags & S_CASEFOLD)
2325 #define IS_VERITY(inode) ((inode)->i_flags & S_VERITY)
2327 #define IS_WHITEOUT(inode) (S_ISCHR(inode->i_mode) && \
2328 (inode)->i_rdev == WHITEOUT_DEV)
2330 static inline bool HAS_UNMAPPED_ID(struct user_namespace *mnt_userns,
2331 struct inode *inode)
2333 return !vfsuid_valid(i_uid_into_vfsuid(mnt_userns, inode)) ||
2334 !vfsgid_valid(i_gid_into_vfsgid(mnt_userns, inode));
2337 static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp)
2339 *kiocb = (struct kiocb) {
2341 .ki_flags = filp->f_iocb_flags,
2342 .ki_ioprio = get_current_ioprio(),
2346 static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
2349 *kiocb = (struct kiocb) {
2351 .ki_flags = kiocb_src->ki_flags,
2352 .ki_ioprio = kiocb_src->ki_ioprio,
2353 .ki_pos = kiocb_src->ki_pos,
2358 * Inode state bits. Protected by inode->i_lock
2360 * Four bits determine the dirty state of the inode: I_DIRTY_SYNC,
2361 * I_DIRTY_DATASYNC, I_DIRTY_PAGES, and I_DIRTY_TIME.
2363 * Four bits define the lifetime of an inode. Initially, inodes are I_NEW,
2364 * until that flag is cleared. I_WILL_FREE, I_FREEING and I_CLEAR are set at
2365 * various stages of removing an inode.
2367 * Two bits are used for locking and completion notification, I_NEW and I_SYNC.
2369 * I_DIRTY_SYNC Inode is dirty, but doesn't have to be written on
2370 * fdatasync() (unless I_DIRTY_DATASYNC is also set).
2371 * Timestamp updates are the usual cause.
2372 * I_DIRTY_DATASYNC Data-related inode changes pending. We keep track of
2373 * these changes separately from I_DIRTY_SYNC so that we
2374 * don't have to write inode on fdatasync() when only
2375 * e.g. the timestamps have changed.
2376 * I_DIRTY_PAGES Inode has dirty pages. Inode itself may be clean.
2377 * I_DIRTY_TIME The inode itself only has dirty timestamps, and the
2378 * lazytime mount option is enabled. We keep track of this
2379 * separately from I_DIRTY_SYNC in order to implement
2380 * lazytime. This gets cleared if I_DIRTY_INODE
2381 * (I_DIRTY_SYNC and/or I_DIRTY_DATASYNC) gets set. I.e.
2382 * either I_DIRTY_TIME *or* I_DIRTY_INODE can be set in
2383 * i_state, but not both. I_DIRTY_PAGES may still be set.
2384 * I_NEW Serves as both a mutex and completion notification.
2385 * New inodes set I_NEW. If two processes both create
2386 * the same inode, one of them will release its inode and
2387 * wait for I_NEW to be released before returning.
2388 * Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can
2389 * also cause waiting on I_NEW, without I_NEW actually
2390 * being set. find_inode() uses this to prevent returning
2391 * nearly-dead inodes.
2392 * I_WILL_FREE Must be set when calling write_inode_now() if i_count
2393 * is zero. I_FREEING must be set when I_WILL_FREE is
2395 * I_FREEING Set when inode is about to be freed but still has dirty
2396 * pages or buffers attached or the inode itself is still
2398 * I_CLEAR Added by clear_inode(). In this state the inode is
2399 * clean and can be destroyed. Inode keeps I_FREEING.
2401 * Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are
2402 * prohibited for many purposes. iget() must wait for
2403 * the inode to be completely released, then create it
2404 * anew. Other functions will just ignore such inodes,
2405 * if appropriate. I_NEW is used for waiting.
2407 * I_SYNC Writeback of inode is running. The bit is set during
2408 * data writeback, and cleared with a wakeup on the bit
2409 * address once it is done. The bit is also used to pin
2410 * the inode in memory for flusher thread.
2412 * I_REFERENCED Marks the inode as recently references on the LRU list.
2414 * I_DIO_WAKEUP Never set. Only used as a key for wait_on_bit().
2416 * I_WB_SWITCH Cgroup bdi_writeback switching in progress. Used to
2417 * synchronize competing switching instances and to tell
2418 * wb stat updates to grab the i_pages lock. See
2419 * inode_switch_wbs_work_fn() for details.
2421 * I_OVL_INUSE Used by overlayfs to get exclusive ownership on upper
2422 * and work dirs among overlayfs mounts.
2424 * I_CREATING New object's inode in the middle of setting up.
2426 * I_DONTCACHE Evict inode as soon as it is not used anymore.
2428 * I_SYNC_QUEUED Inode is queued in b_io or b_more_io writeback lists.
2429 * Used to detect that mark_inode_dirty() should not move
2430 * inode between dirty lists.
2432 * I_PINNING_FSCACHE_WB Inode is pinning an fscache object for writeback.
2434 * Q: What is the difference between I_WILL_FREE and I_FREEING?
2436 #define I_DIRTY_SYNC (1 << 0)
2437 #define I_DIRTY_DATASYNC (1 << 1)
2438 #define I_DIRTY_PAGES (1 << 2)
2440 #define I_NEW (1 << __I_NEW)
2441 #define I_WILL_FREE (1 << 4)
2442 #define I_FREEING (1 << 5)
2443 #define I_CLEAR (1 << 6)
2445 #define I_SYNC (1 << __I_SYNC)
2446 #define I_REFERENCED (1 << 8)
2447 #define __I_DIO_WAKEUP 9
2448 #define I_DIO_WAKEUP (1 << __I_DIO_WAKEUP)
2449 #define I_LINKABLE (1 << 10)
2450 #define I_DIRTY_TIME (1 << 11)
2451 #define I_WB_SWITCH (1 << 13)
2452 #define I_OVL_INUSE (1 << 14)
2453 #define I_CREATING (1 << 15)
2454 #define I_DONTCACHE (1 << 16)
2455 #define I_SYNC_QUEUED (1 << 17)
2456 #define I_PINNING_FSCACHE_WB (1 << 18)
2458 #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC)
2459 #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES)
2460 #define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME)
2462 extern void __mark_inode_dirty(struct inode *, int);
2463 static inline void mark_inode_dirty(struct inode *inode)
2465 __mark_inode_dirty(inode, I_DIRTY);
2468 static inline void mark_inode_dirty_sync(struct inode *inode)
2470 __mark_inode_dirty(inode, I_DIRTY_SYNC);
2474 * Returns true if the given inode itself only has dirty timestamps (its pages
2475 * may still be dirty) and isn't currently being allocated or freed.
2476 * Filesystems should call this if when writing an inode when lazytime is
2477 * enabled, they want to opportunistically write the timestamps of other inodes
2478 * located very nearby on-disk, e.g. in the same inode block. This returns true
2479 * if the given inode is in need of such an opportunistic update. Requires
2480 * i_lock, or at least later re-checking under i_lock.
2482 static inline bool inode_is_dirtytime_only(struct inode *inode)
2484 return (inode->i_state & (I_DIRTY_TIME | I_NEW |
2485 I_FREEING | I_WILL_FREE)) == I_DIRTY_TIME;
2488 extern void inc_nlink(struct inode *inode);
2489 extern void drop_nlink(struct inode *inode);
2490 extern void clear_nlink(struct inode *inode);
2491 extern void set_nlink(struct inode *inode, unsigned int nlink);
2493 static inline void inode_inc_link_count(struct inode *inode)
2496 mark_inode_dirty(inode);
2499 static inline void inode_dec_link_count(struct inode *inode)
2502 mark_inode_dirty(inode);
2505 enum file_time_flags {
2512 extern bool atime_needs_update(const struct path *, struct inode *);
2513 extern void touch_atime(const struct path *);
2514 int inode_update_time(struct inode *inode, struct timespec64 *time, int flags);
2516 static inline void file_accessed(struct file *file)
2518 if (!(file->f_flags & O_NOATIME))
2519 touch_atime(&file->f_path);
2522 extern int file_modified(struct file *file);
2523 int kiocb_modified(struct kiocb *iocb);
2525 int sync_inode_metadata(struct inode *inode, int wait);
2527 struct file_system_type {
2530 #define FS_REQUIRES_DEV 1
2531 #define FS_BINARY_MOUNTDATA 2
2532 #define FS_HAS_SUBTYPE 4
2533 #define FS_USERNS_MOUNT 8 /* Can be mounted by userns root */
2534 #define FS_DISALLOW_NOTIFY_PERM 16 /* Disable fanotify permission events */
2535 #define FS_ALLOW_IDMAP 32 /* FS has been updated to handle vfs idmappings. */
2536 #define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() during rename() internally. */
2537 int (*init_fs_context)(struct fs_context *);
2538 const struct fs_parameter_spec *parameters;
2539 struct dentry *(*mount) (struct file_system_type *, int,
2540 const char *, void *);
2541 void (*kill_sb) (struct super_block *);
2542 struct module *owner;
2543 struct file_system_type * next;
2544 struct hlist_head fs_supers;
2546 struct lock_class_key s_lock_key;
2547 struct lock_class_key s_umount_key;
2548 struct lock_class_key s_vfs_rename_key;
2549 struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
2551 struct lock_class_key i_lock_key;
2552 struct lock_class_key i_mutex_key;
2553 struct lock_class_key invalidate_lock_key;
2554 struct lock_class_key i_mutex_dir_key;
2557 #define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME)
2559 extern struct dentry *mount_bdev(struct file_system_type *fs_type,
2560 int flags, const char *dev_name, void *data,
2561 int (*fill_super)(struct super_block *, void *, int));
2562 extern struct dentry *mount_single(struct file_system_type *fs_type,
2563 int flags, void *data,
2564 int (*fill_super)(struct super_block *, void *, int));
2565 extern struct dentry *mount_nodev(struct file_system_type *fs_type,
2566 int flags, void *data,
2567 int (*fill_super)(struct super_block *, void *, int));
2568 extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path);
2569 void retire_super(struct super_block *sb);
2570 void generic_shutdown_super(struct super_block *sb);
2571 void kill_block_super(struct super_block *sb);
2572 void kill_anon_super(struct super_block *sb);
2573 void kill_litter_super(struct super_block *sb);
2574 void deactivate_super(struct super_block *sb);
2575 void deactivate_locked_super(struct super_block *sb);
2576 int set_anon_super(struct super_block *s, void *data);
2577 int set_anon_super_fc(struct super_block *s, struct fs_context *fc);
2578 int get_anon_bdev(dev_t *);
2579 void free_anon_bdev(dev_t);
2580 struct super_block *sget_fc(struct fs_context *fc,
2581 int (*test)(struct super_block *, struct fs_context *),
2582 int (*set)(struct super_block *, struct fs_context *));
2583 struct super_block *sget(struct file_system_type *type,
2584 int (*test)(struct super_block *,void *),
2585 int (*set)(struct super_block *,void *),
2586 int flags, void *data);
2588 /* Alas, no aliases. Too much hassle with bringing module.h everywhere */
2589 #define fops_get(fops) \
2590 (((fops) && try_module_get((fops)->owner) ? (fops) : NULL))
2591 #define fops_put(fops) \
2592 do { if (fops) module_put((fops)->owner); } while(0)
2594 * This one is to be used *ONLY* from ->open() instances.
2595 * fops must be non-NULL, pinned down *and* module dependencies
2596 * should be sufficient to pin the caller down as well.
2598 #define replace_fops(f, fops) \
2600 struct file *__file = (f); \
2601 fops_put(__file->f_op); \
2602 BUG_ON(!(__file->f_op = (fops))); \
2605 extern int register_filesystem(struct file_system_type *);
2606 extern int unregister_filesystem(struct file_system_type *);
2607 extern int vfs_statfs(const struct path *, struct kstatfs *);
2608 extern int user_statfs(const char __user *, struct kstatfs *);
2609 extern int fd_statfs(int, struct kstatfs *);
2610 extern int freeze_super(struct super_block *super);
2611 extern int thaw_super(struct super_block *super);
2612 extern __printf(2, 3)
2613 int super_setup_bdi_name(struct super_block *sb, char *fmt, ...);
2614 extern int super_setup_bdi(struct super_block *sb);
2616 extern int current_umask(void);
2618 extern void ihold(struct inode * inode);
2619 extern void iput(struct inode *);
2620 extern int generic_update_time(struct inode *, struct timespec64 *, int);
2623 extern struct kobject *fs_kobj;
2625 #define MAX_RW_COUNT (INT_MAX & PAGE_MASK)
2627 #ifdef CONFIG_FILE_LOCKING
2628 static inline int break_lease(struct inode *inode, unsigned int mode)
2631 * Since this check is lockless, we must ensure that any refcounts
2632 * taken are done before checking i_flctx->flc_lease. Otherwise, we
2633 * could end up racing with tasks trying to set a new lease on this
2637 if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease))
2638 return __break_lease(inode, mode, FL_LEASE);
2642 static inline int break_deleg(struct inode *inode, unsigned int mode)
2645 * Since this check is lockless, we must ensure that any refcounts
2646 * taken are done before checking i_flctx->flc_lease. Otherwise, we
2647 * could end up racing with tasks trying to set a new lease on this
2651 if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease))
2652 return __break_lease(inode, mode, FL_DELEG);
2656 static inline int try_break_deleg(struct inode *inode, struct inode **delegated_inode)
2660 ret = break_deleg(inode, O_WRONLY|O_NONBLOCK);
2661 if (ret == -EWOULDBLOCK && delegated_inode) {
2662 *delegated_inode = inode;
2668 static inline int break_deleg_wait(struct inode **delegated_inode)
2672 ret = break_deleg(*delegated_inode, O_WRONLY);
2673 iput(*delegated_inode);
2674 *delegated_inode = NULL;
2678 static inline int break_layout(struct inode *inode, bool wait)
2681 if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease))
2682 return __break_lease(inode,
2683 wait ? O_WRONLY : O_WRONLY | O_NONBLOCK,
2688 #else /* !CONFIG_FILE_LOCKING */
2689 static inline int break_lease(struct inode *inode, unsigned int mode)
2694 static inline int break_deleg(struct inode *inode, unsigned int mode)
2699 static inline int try_break_deleg(struct inode *inode, struct inode **delegated_inode)
2704 static inline int break_deleg_wait(struct inode **delegated_inode)
2710 static inline int break_layout(struct inode *inode, bool wait)
2715 #endif /* CONFIG_FILE_LOCKING */
2720 const char *name; /* pointer to actual string */
2721 const __user char *uptr; /* original userland pointer */
2723 struct audit_names *aname;
2726 static_assert(offsetof(struct filename, iname) % sizeof(long) == 0);
2728 static inline struct user_namespace *file_mnt_user_ns(struct file *file)
2730 return mnt_user_ns(file->f_path.mnt);
2734 * is_idmapped_mnt - check whether a mount is mapped
2735 * @mnt: the mount to check
2737 * If @mnt has an idmapping attached different from the
2738 * filesystem's idmapping then @mnt is mapped.
2740 * Return: true if mount is mapped, false if not.
2742 static inline bool is_idmapped_mnt(const struct vfsmount *mnt)
2744 return mnt_user_ns(mnt) != mnt->mnt_sb->s_user_ns;
2747 extern long vfs_truncate(const struct path *, loff_t);
2748 int do_truncate(struct user_namespace *, struct dentry *, loff_t start,
2749 unsigned int time_attrs, struct file *filp);
2750 extern int vfs_fallocate(struct file *file, int mode, loff_t offset,
2752 extern long do_sys_open(int dfd, const char __user *filename, int flags,
2754 extern struct file *file_open_name(struct filename *, int, umode_t);
2755 extern struct file *filp_open(const char *, int, umode_t);
2756 extern struct file *file_open_root(const struct path *,
2757 const char *, int, umode_t);
2758 static inline struct file *file_open_root_mnt(struct vfsmount *mnt,
2759 const char *name, int flags, umode_t mode)
2761 return file_open_root(&(struct path){.mnt = mnt, .dentry = mnt->mnt_root},
2764 extern struct file * dentry_open(const struct path *, int, const struct cred *);
2765 extern struct file *dentry_create(const struct path *path, int flags,
2766 umode_t mode, const struct cred *cred);
2767 extern struct file * open_with_fake_path(const struct path *, int,
2768 struct inode*, const struct cred *);
2769 static inline struct file *file_clone_open(struct file *file)
2771 return dentry_open(&file->f_path, file->f_flags, file->f_cred);
2773 extern int filp_close(struct file *, fl_owner_t id);
2775 extern struct filename *getname_flags(const char __user *, int, int *);
2776 extern struct filename *getname_uflags(const char __user *, int);
2777 extern struct filename *getname(const char __user *);
2778 extern struct filename *getname_kernel(const char *);
2779 extern void putname(struct filename *name);
2781 extern int finish_open(struct file *file, struct dentry *dentry,
2782 int (*open)(struct inode *, struct file *));
2783 extern int finish_no_open(struct file *file, struct dentry *dentry);
2786 extern void __init vfs_caches_init_early(void);
2787 extern void __init vfs_caches_init(void);
2789 extern struct kmem_cache *names_cachep;
2791 #define __getname() kmem_cache_alloc(names_cachep, GFP_KERNEL)
2792 #define __putname(name) kmem_cache_free(names_cachep, (void *)(name))
2794 extern struct super_block *blockdev_superblock;
2795 static inline bool sb_is_blkdev_sb(struct super_block *sb)
2797 return IS_ENABLED(CONFIG_BLOCK) && sb == blockdev_superblock;
2800 void emergency_thaw_all(void);
2801 extern int sync_filesystem(struct super_block *);
2802 extern const struct file_operations def_blk_fops;
2803 extern const struct file_operations def_chr_fops;
2806 #define CHRDEV_MAJOR_MAX 512
2807 /* Marks the bottom of the first segment of free char majors */
2808 #define CHRDEV_MAJOR_DYN_END 234
2809 /* Marks the top and bottom of the second segment of free char majors */
2810 #define CHRDEV_MAJOR_DYN_EXT_START 511
2811 #define CHRDEV_MAJOR_DYN_EXT_END 384
2813 extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *);
2814 extern int register_chrdev_region(dev_t, unsigned, const char *);
2815 extern int __register_chrdev(unsigned int major, unsigned int baseminor,
2816 unsigned int count, const char *name,
2817 const struct file_operations *fops);
2818 extern void __unregister_chrdev(unsigned int major, unsigned int baseminor,
2819 unsigned int count, const char *name);
2820 extern void unregister_chrdev_region(dev_t, unsigned);
2821 extern void chrdev_show(struct seq_file *,off_t);
2823 static inline int register_chrdev(unsigned int major, const char *name,
2824 const struct file_operations *fops)
2826 return __register_chrdev(major, 0, 256, name, fops);
2829 static inline void unregister_chrdev(unsigned int major, const char *name)
2831 __unregister_chrdev(major, 0, 256, name);
2834 extern void init_special_inode(struct inode *, umode_t, dev_t);
2836 /* Invalid inode operations -- fs/bad_inode.c */
2837 extern void make_bad_inode(struct inode *);
2838 extern bool is_bad_inode(struct inode *);
2840 extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart,
2842 extern int __must_check file_check_and_advance_wb_err(struct file *file);
2843 extern int __must_check file_write_and_wait_range(struct file *file,
2844 loff_t start, loff_t end);
2846 static inline int file_write_and_wait(struct file *file)
2848 return file_write_and_wait_range(file, 0, LLONG_MAX);
2851 extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end,
2853 extern int vfs_fsync(struct file *file, int datasync);
2855 extern int sync_file_range(struct file *file, loff_t offset, loff_t nbytes,
2856 unsigned int flags);
2858 static inline bool iocb_is_dsync(const struct kiocb *iocb)
2860 return (iocb->ki_flags & IOCB_DSYNC) ||
2861 IS_SYNC(iocb->ki_filp->f_mapping->host);
2865 * Sync the bytes written if this was a synchronous write. Expect ki_pos
2866 * to already be updated for the write, and will return either the amount
2867 * of bytes passed in, or an error if syncing the file failed.
2869 static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count)
2871 if (iocb_is_dsync(iocb)) {
2872 int ret = vfs_fsync_range(iocb->ki_filp,
2873 iocb->ki_pos - count, iocb->ki_pos - 1,
2874 (iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
2882 extern void emergency_sync(void);
2883 extern void emergency_remount(void);
2886 extern int bmap(struct inode *inode, sector_t *block);
2888 static inline int bmap(struct inode *inode, sector_t *block)
2894 int notify_change(struct user_namespace *, struct dentry *,
2895 struct iattr *, struct inode **);
2896 int inode_permission(struct user_namespace *, struct inode *, int);
2897 int generic_permission(struct user_namespace *, struct inode *, int);
2898 static inline int file_permission(struct file *file, int mask)
2900 return inode_permission(file_mnt_user_ns(file),
2901 file_inode(file), mask);
2903 static inline int path_permission(const struct path *path, int mask)
2905 return inode_permission(mnt_user_ns(path->mnt),
2906 d_inode(path->dentry), mask);
2908 int __check_sticky(struct user_namespace *mnt_userns, struct inode *dir,
2909 struct inode *inode);
2911 static inline bool execute_ok(struct inode *inode)
2913 return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode);
2916 static inline bool inode_wrong_type(const struct inode *inode, umode_t mode)
2918 return (inode->i_mode ^ mode) & S_IFMT;
2921 static inline void file_start_write(struct file *file)
2923 if (!S_ISREG(file_inode(file)->i_mode))
2925 sb_start_write(file_inode(file)->i_sb);
2928 static inline bool file_start_write_trylock(struct file *file)
2930 if (!S_ISREG(file_inode(file)->i_mode))
2932 return sb_start_write_trylock(file_inode(file)->i_sb);
2935 static inline void file_end_write(struct file *file)
2937 if (!S_ISREG(file_inode(file)->i_mode))
2939 __sb_end_write(file_inode(file)->i_sb, SB_FREEZE_WRITE);
2943 * This is used for regular files where some users -- especially the
2944 * currently executed binary in a process, previously handled via
2945 * VM_DENYWRITE -- cannot handle concurrent write (and maybe mmap
2946 * read-write shared) accesses.
2948 * get_write_access() gets write permission for a file.
2949 * put_write_access() releases this write permission.
2950 * deny_write_access() denies write access to a file.
2951 * allow_write_access() re-enables write access to a file.
2953 * The i_writecount field of an inode can have the following values:
2954 * 0: no write access, no denied write access
2955 * < 0: (-i_writecount) users that denied write access to the file.
2956 * > 0: (i_writecount) users that have write access to the file.
2958 * Normally we operate on that counter with atomic_{inc,dec} and it's safe
2959 * except for the cases where we don't hold i_writecount yet. Then we need to
2960 * use {get,deny}_write_access() - these functions check the sign and refuse
2961 * to do the change if sign is wrong.
2963 static inline int get_write_access(struct inode *inode)
2965 return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY;
2967 static inline int deny_write_access(struct file *file)
2969 struct inode *inode = file_inode(file);
2970 return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY;
2972 static inline void put_write_access(struct inode * inode)
2974 atomic_dec(&inode->i_writecount);
2976 static inline void allow_write_access(struct file *file)
2979 atomic_inc(&file_inode(file)->i_writecount);
2981 static inline bool inode_is_open_for_write(const struct inode *inode)
2983 return atomic_read(&inode->i_writecount) > 0;
2986 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
2987 static inline void i_readcount_dec(struct inode *inode)
2989 BUG_ON(!atomic_read(&inode->i_readcount));
2990 atomic_dec(&inode->i_readcount);
2992 static inline void i_readcount_inc(struct inode *inode)
2994 atomic_inc(&inode->i_readcount);
2997 static inline void i_readcount_dec(struct inode *inode)
3001 static inline void i_readcount_inc(struct inode *inode)
3006 extern int do_pipe_flags(int *, int);
3008 extern ssize_t kernel_read(struct file *, void *, size_t, loff_t *);
3009 ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos);
3010 extern ssize_t kernel_write(struct file *, const void *, size_t, loff_t *);
3011 extern ssize_t __kernel_write(struct file *, const void *, size_t, loff_t *);
3012 extern struct file * open_exec(const char *);
3014 /* fs/dcache.c -- generic fs support functions */
3015 extern bool is_subdir(struct dentry *, struct dentry *);
3016 extern bool path_is_under(const struct path *, const struct path *);
3018 extern char *file_path(struct file *, char *, int);
3020 #include <linux/err.h>
3022 /* needed for stackable file system support */
3023 extern loff_t default_llseek(struct file *file, loff_t offset, int whence);
3025 extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence);
3027 extern int inode_init_always(struct super_block *, struct inode *);
3028 extern void inode_init_once(struct inode *);
3029 extern void address_space_init_once(struct address_space *mapping);
3030 extern struct inode * igrab(struct inode *);
3031 extern ino_t iunique(struct super_block *, ino_t);
3032 extern int inode_needs_sync(struct inode *inode);
3033 extern int generic_delete_inode(struct inode *inode);
3034 static inline int generic_drop_inode(struct inode *inode)
3036 return !inode->i_nlink || inode_unhashed(inode);
3038 extern void d_mark_dontcache(struct inode *inode);
3040 extern struct inode *ilookup5_nowait(struct super_block *sb,
3041 unsigned long hashval, int (*test)(struct inode *, void *),
3043 extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
3044 int (*test)(struct inode *, void *), void *data);
3045 extern struct inode *ilookup(struct super_block *sb, unsigned long ino);
3047 extern struct inode *inode_insert5(struct inode *inode, unsigned long hashval,
3048 int (*test)(struct inode *, void *),
3049 int (*set)(struct inode *, void *),
3051 extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *);
3052 extern struct inode * iget_locked(struct super_block *, unsigned long);
3053 extern struct inode *find_inode_nowait(struct super_block *,
3055 int (*match)(struct inode *,
3056 unsigned long, void *),
3058 extern struct inode *find_inode_rcu(struct super_block *, unsigned long,
3059 int (*)(struct inode *, void *), void *);
3060 extern struct inode *find_inode_by_ino_rcu(struct super_block *, unsigned long);
3061 extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *);
3062 extern int insert_inode_locked(struct inode *);
3063 #ifdef CONFIG_DEBUG_LOCK_ALLOC
3064 extern void lockdep_annotate_inode_mutex_key(struct inode *inode);
3066 static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { };
3068 extern void unlock_new_inode(struct inode *);
3069 extern void discard_new_inode(struct inode *);
3070 extern unsigned int get_next_ino(void);
3071 extern void evict_inodes(struct super_block *sb);
3072 void dump_mapping(const struct address_space *);
3075 * Userspace may rely on the the inode number being non-zero. For example, glibc
3076 * simply ignores files with zero i_ino in unlink() and other places.
3078 * As an additional complication, if userspace was compiled with
3079 * _FILE_OFFSET_BITS=32 on a 64-bit kernel we'll only end up reading out the
3080 * lower 32 bits, so we need to check that those aren't zero explicitly. With
3081 * _FILE_OFFSET_BITS=64, this may cause some harmless false-negatives, but
3082 * better safe than sorry.
3084 static inline bool is_zero_ino(ino_t ino)
3086 return (u32)ino == 0;
3089 extern void __iget(struct inode * inode);
3090 extern void iget_failed(struct inode *);
3091 extern void clear_inode(struct inode *);
3092 extern void __destroy_inode(struct inode *);
3093 extern struct inode *new_inode_pseudo(struct super_block *sb);
3094 extern struct inode *new_inode(struct super_block *sb);
3095 extern void free_inode_nonrcu(struct inode *inode);
3096 extern int should_remove_suid(struct dentry *);
3097 extern int file_remove_privs(struct file *);
3100 * This must be used for allocating filesystems specific inodes to set
3101 * up the inode reclaim context correctly.
3103 static inline void *
3104 alloc_inode_sb(struct super_block *sb, struct kmem_cache *cache, gfp_t gfp)
3106 return kmem_cache_alloc_lru(cache, &sb->s_inode_lru, gfp);
3109 extern void __insert_inode_hash(struct inode *, unsigned long hashval);
3110 static inline void insert_inode_hash(struct inode *inode)
3112 __insert_inode_hash(inode, inode->i_ino);
3115 extern void __remove_inode_hash(struct inode *);
3116 static inline void remove_inode_hash(struct inode *inode)
3118 if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash))
3119 __remove_inode_hash(inode);
3122 extern void inode_sb_list_add(struct inode *inode);
3123 extern void inode_add_lru(struct inode *inode);
3125 extern int sb_set_blocksize(struct super_block *, int);
3126 extern int sb_min_blocksize(struct super_block *, int);
3128 extern int generic_file_mmap(struct file *, struct vm_area_struct *);
3129 extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *);
3130 extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *);
3131 int generic_write_checks_count(struct kiocb *iocb, loff_t *count);
3132 extern int generic_write_check_limits(struct file *file, loff_t pos,
3134 extern int generic_file_rw_checks(struct file *file_in, struct file *file_out);
3135 ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *to,
3136 ssize_t already_read);
3137 extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *);
3138 extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *);
3139 extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *);
3140 extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *);
3141 ssize_t generic_perform_write(struct kiocb *, struct iov_iter *);
3143 ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos,
3145 ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos,
3147 ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb,
3148 struct iov_iter *iter);
3149 ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb,
3150 struct iov_iter *iter);
3153 extern ssize_t generic_file_splice_read(struct file *, loff_t *,
3154 struct pipe_inode_info *, size_t, unsigned int);
3155 extern ssize_t iter_file_splice_write(struct pipe_inode_info *,
3156 struct file *, loff_t *, size_t, unsigned int);
3157 extern ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe,
3158 struct file *out, loff_t *, size_t len, unsigned int flags);
3159 extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
3160 loff_t *opos, size_t len, unsigned int flags);
3164 file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping);
3165 extern loff_t noop_llseek(struct file *file, loff_t offset, int whence);
3166 #define no_llseek NULL
3167 extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize);
3168 extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence);
3169 extern loff_t generic_file_llseek_size(struct file *file, loff_t offset,
3170 int whence, loff_t maxsize, loff_t eof);
3171 extern loff_t fixed_size_llseek(struct file *file, loff_t offset,
3172 int whence, loff_t size);
3173 extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t);
3174 extern loff_t no_seek_end_llseek(struct file *, loff_t, int);
3175 int rw_verify_area(int, struct file *, const loff_t *, size_t);
3176 extern int generic_file_open(struct inode * inode, struct file * filp);
3177 extern int nonseekable_open(struct inode * inode, struct file * filp);
3178 extern int stream_open(struct inode * inode, struct file * filp);
3181 typedef void (dio_submit_t)(struct bio *bio, struct inode *inode,
3182 loff_t file_offset);
3185 /* need locking between buffered and direct access */
3188 /* filesystem does not support filling holes */
3189 DIO_SKIP_HOLES = 0x02,
3192 ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
3193 struct block_device *bdev, struct iov_iter *iter,
3194 get_block_t get_block,
3195 dio_iodone_t end_io, dio_submit_t submit_io,
3198 static inline ssize_t blockdev_direct_IO(struct kiocb *iocb,
3199 struct inode *inode,
3200 struct iov_iter *iter,
3201 get_block_t get_block)
3203 return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter,
3204 get_block, NULL, NULL, DIO_LOCKING | DIO_SKIP_HOLES);
3208 void inode_dio_wait(struct inode *inode);
3211 * inode_dio_begin - signal start of a direct I/O requests
3212 * @inode: inode the direct I/O happens on
3214 * This is called once we've finished processing a direct I/O request,
3215 * and is used to wake up callers waiting for direct I/O to be quiesced.
3217 static inline void inode_dio_begin(struct inode *inode)
3219 atomic_inc(&inode->i_dio_count);
3223 * inode_dio_end - signal finish of a direct I/O requests
3224 * @inode: inode the direct I/O happens on
3226 * This is called once we've finished processing a direct I/O request,
3227 * and is used to wake up callers waiting for direct I/O to be quiesced.
3229 static inline void inode_dio_end(struct inode *inode)
3231 if (atomic_dec_and_test(&inode->i_dio_count))
3232 wake_up_bit(&inode->i_state, __I_DIO_WAKEUP);
3236 * Warn about a page cache invalidation failure diring a direct I/O write.
3238 void dio_warn_stale_pagecache(struct file *filp);
3240 extern void inode_set_flags(struct inode *inode, unsigned int flags,
3243 extern const struct file_operations generic_ro_fops;
3245 #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m))
3247 extern int readlink_copy(char __user *, int, const char *);
3248 extern int page_readlink(struct dentry *, char __user *, int);
3249 extern const char *page_get_link(struct dentry *, struct inode *,
3250 struct delayed_call *);
3251 extern void page_put_link(void *);
3252 extern int page_symlink(struct inode *inode, const char *symname, int len);
3253 extern const struct inode_operations page_symlink_inode_operations;
3254 extern void kfree_link(void *);
3255 void generic_fillattr(struct user_namespace *, struct inode *, struct kstat *);
3256 void generic_fill_statx_attr(struct inode *inode, struct kstat *stat);
3257 extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int);
3258 extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int);
3259 void __inode_add_bytes(struct inode *inode, loff_t bytes);
3260 void inode_add_bytes(struct inode *inode, loff_t bytes);
3261 void __inode_sub_bytes(struct inode *inode, loff_t bytes);
3262 void inode_sub_bytes(struct inode *inode, loff_t bytes);
3263 static inline loff_t __inode_get_bytes(struct inode *inode)
3265 return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes;
3267 loff_t inode_get_bytes(struct inode *inode);
3268 void inode_set_bytes(struct inode *inode, loff_t bytes);
3269 const char *simple_get_link(struct dentry *, struct inode *,
3270 struct delayed_call *);
3271 extern const struct inode_operations simple_symlink_inode_operations;
3273 extern int iterate_dir(struct file *, struct dir_context *);
3275 int vfs_fstatat(int dfd, const char __user *filename, struct kstat *stat,
3277 int vfs_fstat(int fd, struct kstat *stat);
3279 static inline int vfs_stat(const char __user *filename, struct kstat *stat)
3281 return vfs_fstatat(AT_FDCWD, filename, stat, 0);
3283 static inline int vfs_lstat(const char __user *name, struct kstat *stat)
3285 return vfs_fstatat(AT_FDCWD, name, stat, AT_SYMLINK_NOFOLLOW);
3288 extern const char *vfs_get_link(struct dentry *, struct delayed_call *);
3289 extern int vfs_readlink(struct dentry *, char __user *, int);
3291 extern struct file_system_type *get_filesystem(struct file_system_type *fs);
3292 extern void put_filesystem(struct file_system_type *fs);
3293 extern struct file_system_type *get_fs_type(const char *name);
3294 extern struct super_block *get_super(struct block_device *);
3295 extern struct super_block *get_active_super(struct block_device *bdev);
3296 extern void drop_super(struct super_block *sb);
3297 extern void drop_super_exclusive(struct super_block *sb);
3298 extern void iterate_supers(void (*)(struct super_block *, void *), void *);
3299 extern void iterate_supers_type(struct file_system_type *,
3300 void (*)(struct super_block *, void *), void *);
3302 extern int dcache_dir_open(struct inode *, struct file *);
3303 extern int dcache_dir_close(struct inode *, struct file *);
3304 extern loff_t dcache_dir_lseek(struct file *, loff_t, int);
3305 extern int dcache_readdir(struct file *, struct dir_context *);
3306 extern int simple_setattr(struct user_namespace *, struct dentry *,
3308 extern int simple_getattr(struct user_namespace *, const struct path *,
3309 struct kstat *, u32, unsigned int);
3310 extern int simple_statfs(struct dentry *, struct kstatfs *);
3311 extern int simple_open(struct inode *inode, struct file *file);
3312 extern int simple_link(struct dentry *, struct inode *, struct dentry *);
3313 extern int simple_unlink(struct inode *, struct dentry *);
3314 extern int simple_rmdir(struct inode *, struct dentry *);
3315 extern int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry,
3316 struct inode *new_dir, struct dentry *new_dentry);
3317 extern int simple_rename(struct user_namespace *, struct inode *,
3318 struct dentry *, struct inode *, struct dentry *,
3320 extern void simple_recursive_removal(struct dentry *,
3321 void (*callback)(struct dentry *));
3322 extern int noop_fsync(struct file *, loff_t, loff_t, int);
3323 extern ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter);
3324 extern int simple_empty(struct dentry *);
3325 extern int simple_write_begin(struct file *file, struct address_space *mapping,
3326 loff_t pos, unsigned len,
3327 struct page **pagep, void **fsdata);
3328 extern const struct address_space_operations ram_aops;
3329 extern int always_delete_dentry(const struct dentry *);
3330 extern struct inode *alloc_anon_inode(struct super_block *);
3331 extern int simple_nosetlease(struct file *, long, struct file_lock **, void **);
3332 extern const struct dentry_operations simple_dentry_operations;
3334 extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags);
3335 extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *);
3336 extern const struct file_operations simple_dir_operations;
3337 extern const struct inode_operations simple_dir_inode_operations;
3338 extern void make_empty_dir_inode(struct inode *inode);
3339 extern bool is_empty_dir_inode(struct inode *inode);
3340 struct tree_descr { const char *name; const struct file_operations *ops; int mode; };
3341 struct dentry *d_alloc_name(struct dentry *, const char *);
3342 extern int simple_fill_super(struct super_block *, unsigned long,
3343 const struct tree_descr *);
3344 extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count);
3345 extern void simple_release_fs(struct vfsmount **mount, int *count);
3347 extern ssize_t simple_read_from_buffer(void __user *to, size_t count,
3348 loff_t *ppos, const void *from, size_t available);
3349 extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
3350 const void __user *from, size_t count);
3352 extern int __generic_file_fsync(struct file *, loff_t, loff_t, int);
3353 extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
3355 extern int generic_check_addressable(unsigned, u64);
3357 extern void generic_set_encrypted_ci_d_ops(struct dentry *dentry);
3359 int may_setattr(struct user_namespace *mnt_userns, struct inode *inode,
3360 unsigned int ia_valid);
3361 int setattr_prepare(struct user_namespace *, struct dentry *, struct iattr *);
3362 extern int inode_newsize_ok(const struct inode *, loff_t offset);
3363 void setattr_copy(struct user_namespace *, struct inode *inode,
3364 const struct iattr *attr);
3366 extern int file_update_time(struct file *file);
3368 static inline bool vma_is_dax(const struct vm_area_struct *vma)
3370 return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host);
3373 static inline bool vma_is_fsdax(struct vm_area_struct *vma)
3375 struct inode *inode;
3377 if (!IS_ENABLED(CONFIG_FS_DAX) || !vma->vm_file)
3379 if (!vma_is_dax(vma))
3381 inode = file_inode(vma->vm_file);
3382 if (S_ISCHR(inode->i_mode))
3383 return false; /* device-dax */
3387 static inline int iocb_flags(struct file *file)
3390 if (file->f_flags & O_APPEND)
3392 if (file->f_flags & O_DIRECT)
3394 if (file->f_flags & O_DSYNC)
3396 if (file->f_flags & __O_SYNC)
3401 static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags)
3403 int kiocb_flags = 0;
3405 /* make sure there's no overlap between RWF and private IOCB flags */
3406 BUILD_BUG_ON((__force int) RWF_SUPPORTED & IOCB_EVENTFD);
3410 if (unlikely(flags & ~RWF_SUPPORTED))
3413 if (flags & RWF_NOWAIT) {
3414 if (!(ki->ki_filp->f_mode & FMODE_NOWAIT))
3416 kiocb_flags |= IOCB_NOIO;
3418 kiocb_flags |= (__force int) (flags & RWF_SUPPORTED);
3419 if (flags & RWF_SYNC)
3420 kiocb_flags |= IOCB_DSYNC;
3422 ki->ki_flags |= kiocb_flags;
3426 static inline ino_t parent_ino(struct dentry *dentry)
3431 * Don't strictly need d_lock here? If the parent ino could change
3432 * then surely we'd have a deeper race in the caller?
3434 spin_lock(&dentry->d_lock);
3435 res = dentry->d_parent->d_inode->i_ino;
3436 spin_unlock(&dentry->d_lock);
3440 /* Transaction based IO helpers */
3443 * An argresp is stored in an allocated page and holds the
3444 * size of the argument or response, along with its content
3446 struct simple_transaction_argresp {
3451 #define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp))
3453 char *simple_transaction_get(struct file *file, const char __user *buf,
3455 ssize_t simple_transaction_read(struct file *file, char __user *buf,
3456 size_t size, loff_t *pos);
3457 int simple_transaction_release(struct inode *inode, struct file *file);
3459 void simple_transaction_set(struct file *file, size_t n);
3462 * simple attribute files
3464 * These attributes behave similar to those in sysfs:
3466 * Writing to an attribute immediately sets a value, an open file can be
3467 * written to multiple times.
3469 * Reading from an attribute creates a buffer from the value that might get
3470 * read with multiple read calls. When the attribute has been read
3471 * completely, no further read calls are possible until the file is opened
3474 * All attributes contain a text representation of a numeric value
3475 * that are accessed with the get() and set() functions.
3477 #define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt) \
3478 static int __fops ## _open(struct inode *inode, struct file *file) \
3480 __simple_attr_check_format(__fmt, 0ull); \
3481 return simple_attr_open(inode, file, __get, __set, __fmt); \
3483 static const struct file_operations __fops = { \
3484 .owner = THIS_MODULE, \
3485 .open = __fops ## _open, \
3486 .release = simple_attr_release, \
3487 .read = simple_attr_read, \
3488 .write = simple_attr_write, \
3489 .llseek = generic_file_llseek, \
3492 static inline __printf(1, 2)
3493 void __simple_attr_check_format(const char *fmt, ...)
3495 /* don't do anything, just let the compiler check the arguments; */
3498 int simple_attr_open(struct inode *inode, struct file *file,
3499 int (*get)(void *, u64 *), int (*set)(void *, u64),
3501 int simple_attr_release(struct inode *inode, struct file *file);
3502 ssize_t simple_attr_read(struct file *file, char __user *buf,
3503 size_t len, loff_t *ppos);
3504 ssize_t simple_attr_write(struct file *file, const char __user *buf,
3505 size_t len, loff_t *ppos);
3508 int __init list_bdev_fs_names(char *buf, size_t size);
3510 #define __FMODE_EXEC ((__force int) FMODE_EXEC)
3511 #define __FMODE_NONOTIFY ((__force int) FMODE_NONOTIFY)
3513 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
3514 #define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \
3515 (flag & __FMODE_NONOTIFY)))
3517 static inline bool is_sxid(umode_t mode)
3519 return (mode & S_ISUID) || ((mode & S_ISGID) && (mode & S_IXGRP));
3522 static inline int check_sticky(struct user_namespace *mnt_userns,
3523 struct inode *dir, struct inode *inode)
3525 if (!(dir->i_mode & S_ISVTX))
3528 return __check_sticky(mnt_userns, dir, inode);
3531 static inline void inode_has_no_xattr(struct inode *inode)
3533 if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC))
3534 inode->i_flags |= S_NOSEC;
3537 static inline bool is_root_inode(struct inode *inode)
3539 return inode == inode->i_sb->s_root->d_inode;
3542 static inline bool dir_emit(struct dir_context *ctx,
3543 const char *name, int namelen,
3544 u64 ino, unsigned type)
3546 return ctx->actor(ctx, name, namelen, ctx->pos, ino, type) == 0;
3548 static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx)
3550 return ctx->actor(ctx, ".", 1, ctx->pos,
3551 file->f_path.dentry->d_inode->i_ino, DT_DIR) == 0;
3553 static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx)
3555 return ctx->actor(ctx, "..", 2, ctx->pos,
3556 parent_ino(file->f_path.dentry), DT_DIR) == 0;
3558 static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx)
3560 if (ctx->pos == 0) {
3561 if (!dir_emit_dot(file, ctx))
3565 if (ctx->pos == 1) {
3566 if (!dir_emit_dotdot(file, ctx))
3572 static inline bool dir_relax(struct inode *inode)
3574 inode_unlock(inode);
3576 return !IS_DEADDIR(inode);
3579 static inline bool dir_relax_shared(struct inode *inode)
3581 inode_unlock_shared(inode);
3582 inode_lock_shared(inode);
3583 return !IS_DEADDIR(inode);
3586 extern bool path_noexec(const struct path *path);
3587 extern void inode_nohighmem(struct inode *inode);
3590 extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len,
3592 extern int generic_fadvise(struct file *file, loff_t offset, loff_t len,
3595 #endif /* _LINUX_FS_H */