1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _LINUX_FS_H 3 #define _LINUX_FS_H 4 5 #include <linux/linkage.h> 6 #include <linux/wait_bit.h> 7 #include <linux/kdev_t.h> 8 #include <linux/dcache.h> 9 #include <linux/path.h> 10 #include <linux/stat.h> 11 #include <linux/cache.h> 12 #include <linux/list.h> 13 #include <linux/list_lru.h> 14 #include <linux/llist.h> 15 #include <linux/radix-tree.h> 16 #include <linux/xarray.h> 17 #include <linux/rbtree.h> 18 #include <linux/init.h> 19 #include <linux/pid.h> 20 #include <linux/bug.h> 21 #include <linux/mutex.h> 22 #include <linux/rwsem.h> 23 #include <linux/mm_types.h> 24 #include <linux/capability.h> 25 #include <linux/semaphore.h> 26 #include <linux/fcntl.h> 27 #include <linux/rculist_bl.h> 28 #include <linux/atomic.h> 29 #include <linux/shrinker.h> 30 #include <linux/migrate_mode.h> 31 #include <linux/uidgid.h> 32 #include <linux/lockdep.h> 33 #include <linux/percpu-rwsem.h> 34 #include <linux/workqueue.h> 35 #include <linux/delayed_call.h> 36 #include <linux/uuid.h> 37 #include <linux/errseq.h> 38 #include <linux/ioprio.h> 39 #include <linux/fs_types.h> 40 #include <linux/build_bug.h> 41 #include <linux/stddef.h> 42 #include <linux/mount.h> 43 #include <linux/cred.h> 44 #include <linux/mnt_idmapping.h> 45 #include <linux/slab.h> 46 47 #include <asm/byteorder.h> 48 #include <uapi/linux/fs.h> 49 50 struct backing_dev_info; 51 struct bdi_writeback; 52 struct bio; 53 struct io_comp_batch; 54 struct export_operations; 55 struct fiemap_extent_info; 56 struct hd_geometry; 57 struct iovec; 58 struct kiocb; 59 struct kobject; 60 struct pipe_inode_info; 61 struct poll_table_struct; 62 struct kstatfs; 63 struct vm_area_struct; 64 struct vfsmount; 65 struct cred; 66 struct swap_info_struct; 67 struct seq_file; 68 struct workqueue_struct; 69 struct iov_iter; 70 struct fscrypt_info; 71 struct fscrypt_operations; 72 struct fsverity_info; 73 struct fsverity_operations; 74 struct fs_context; 75 struct fs_parameter_spec; 76 struct fileattr; 77 struct iomap_ops; 78 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); 83 84 extern unsigned long get_max_files(void); 85 extern unsigned int sysctl_nr_open; 86 87 typedef __kernel_rwf_t rwf_t; 88 89 struct buffer_head; 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); 94 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 104 105 /* 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() 108 */ 109 110 /* file is open for reading */ 111 #define FMODE_READ ((__force fmode_t)0x1) 112 /* file is open for writing */ 113 #define FMODE_WRITE ((__force fmode_t)0x2) 114 /* file is seekable */ 115 #define FMODE_LSEEK ((__force fmode_t)0x4) 116 /* file can be accessed using pread */ 117 #define FMODE_PREAD ((__force fmode_t)0x8) 118 /* file can be accessed using pwrite */ 119 #define FMODE_PWRITE ((__force fmode_t)0x10) 120 /* File is opened for execution with sys_execve / sys_uselib */ 121 #define FMODE_EXEC ((__force fmode_t)0x20) 122 /* 32bit hashes as llseek() offset (for directories) */ 123 #define FMODE_32BITHASH ((__force fmode_t)0x200) 124 /* 64bit hashes as llseek() offset (for directories) */ 125 #define FMODE_64BITHASH ((__force fmode_t)0x400) 126 127 /* 128 * Don't update ctime and mtime. 129 * 130 * Currently a special hack for the XFS open_by_handle ioctl, but we'll 131 * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon. 132 */ 133 #define FMODE_NOCMTIME ((__force fmode_t)0x800) 134 135 /* Expect random access pattern */ 136 #define FMODE_RANDOM ((__force fmode_t)0x1000) 137 138 /* File is huge (eg. /dev/mem): treat loff_t as unsigned */ 139 #define FMODE_UNSIGNED_OFFSET ((__force fmode_t)0x2000) 140 141 /* File is opened with O_PATH; almost nothing can be done with it */ 142 #define FMODE_PATH ((__force fmode_t)0x4000) 143 144 /* File needs atomic accesses to f_pos */ 145 #define FMODE_ATOMIC_POS ((__force fmode_t)0x8000) 146 /* Write access to underlying fs */ 147 #define FMODE_WRITER ((__force fmode_t)0x10000) 148 /* Has read method(s) */ 149 #define FMODE_CAN_READ ((__force fmode_t)0x20000) 150 /* Has write method(s) */ 151 #define FMODE_CAN_WRITE ((__force fmode_t)0x40000) 152 153 #define FMODE_OPENED ((__force fmode_t)0x80000) 154 #define FMODE_CREATED ((__force fmode_t)0x100000) 155 156 /* File is stream-like */ 157 #define FMODE_STREAM ((__force fmode_t)0x200000) 158 159 /* File supports DIRECT IO */ 160 #define FMODE_CAN_ODIRECT ((__force fmode_t)0x400000) 161 162 #define FMODE_NOREUSE ((__force fmode_t)0x800000) 163 164 /* File supports non-exclusive O_DIRECT writes from multiple threads */ 165 #define FMODE_DIO_PARALLEL_WRITE ((__force fmode_t)0x1000000) 166 167 /* File is embedded in backing_file object */ 168 #define FMODE_BACKING ((__force fmode_t)0x2000000) 169 170 /* File was opened by fanotify and shouldn't generate fanotify events */ 171 #define FMODE_NONOTIFY ((__force fmode_t)0x4000000) 172 173 /* File is capable of returning -EAGAIN if I/O will block */ 174 #define FMODE_NOWAIT ((__force fmode_t)0x8000000) 175 176 /* File represents mount that needs unmounting */ 177 #define FMODE_NEED_UNMOUNT ((__force fmode_t)0x10000000) 178 179 /* File does not contribute to nr_files count */ 180 #define FMODE_NOACCOUNT ((__force fmode_t)0x20000000) 181 182 /* File supports async buffered reads */ 183 #define FMODE_BUF_RASYNC ((__force fmode_t)0x40000000) 184 185 /* File supports async nowait buffered writes */ 186 #define FMODE_BUF_WASYNC ((__force fmode_t)0x80000000) 187 188 /* 189 * Attribute flags. These should be or-ed together to figure out what 190 * has been changed! 191 */ 192 #define ATTR_MODE (1 << 0) 193 #define ATTR_UID (1 << 1) 194 #define ATTR_GID (1 << 2) 195 #define ATTR_SIZE (1 << 3) 196 #define ATTR_ATIME (1 << 4) 197 #define ATTR_MTIME (1 << 5) 198 #define ATTR_CTIME (1 << 6) 199 #define ATTR_ATIME_SET (1 << 7) 200 #define ATTR_MTIME_SET (1 << 8) 201 #define ATTR_FORCE (1 << 9) /* Not a change, but a change it */ 202 #define ATTR_KILL_SUID (1 << 11) 203 #define ATTR_KILL_SGID (1 << 12) 204 #define ATTR_FILE (1 << 13) 205 #define ATTR_KILL_PRIV (1 << 14) 206 #define ATTR_OPEN (1 << 15) /* Truncating from open(O_TRUNC) */ 207 #define ATTR_TIMES_SET (1 << 16) 208 #define ATTR_TOUCH (1 << 17) 209 210 /* 211 * Whiteout is represented by a char device. The following constants define the 212 * mode and device number to use. 213 */ 214 #define WHITEOUT_MODE 0 215 #define WHITEOUT_DEV 0 216 217 /* 218 * This is the Inode Attributes structure, used for notify_change(). It 219 * uses the above definitions as flags, to know which values have changed. 220 * Also, in this manner, a Filesystem can look at only the values it cares 221 * about. Basically, these are the attributes that the VFS layer can 222 * request to change from the FS layer. 223 * 224 * Derek Atkins <warlord@MIT.EDU> 94-10-20 225 */ 226 struct iattr { 227 unsigned int ia_valid; 228 umode_t ia_mode; 229 /* 230 * The two anonymous unions wrap structures with the same member. 231 * 232 * Filesystems raising FS_ALLOW_IDMAP need to use ia_vfs{g,u}id which 233 * are a dedicated type requiring the filesystem to use the dedicated 234 * helpers. Other filesystem can continue to use ia_{g,u}id until they 235 * have been ported. 236 * 237 * They always contain the same value. In other words FS_ALLOW_IDMAP 238 * pass down the same value on idmapped mounts as they would on regular 239 * mounts. 240 */ 241 union { 242 kuid_t ia_uid; 243 vfsuid_t ia_vfsuid; 244 }; 245 union { 246 kgid_t ia_gid; 247 vfsgid_t ia_vfsgid; 248 }; 249 loff_t ia_size; 250 struct timespec64 ia_atime; 251 struct timespec64 ia_mtime; 252 struct timespec64 ia_ctime; 253 254 /* 255 * Not an attribute, but an auxiliary info for filesystems wanting to 256 * implement an ftruncate() like method. NOTE: filesystem should 257 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL). 258 */ 259 struct file *ia_file; 260 }; 261 262 /* 263 * Includes for diskquotas. 264 */ 265 #include <linux/quota.h> 266 267 /* 268 * Maximum number of layers of fs stack. Needs to be limited to 269 * prevent kernel stack overflow 270 */ 271 #define FILESYSTEM_MAX_STACK_DEPTH 2 272 273 /** 274 * enum positive_aop_returns - aop return codes with specific semantics 275 * 276 * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has 277 * completed, that the page is still locked, and 278 * should be considered active. The VM uses this hint 279 * to return the page to the active list -- it won't 280 * be a candidate for writeback again in the near 281 * future. Other callers must be careful to unlock 282 * the page if they get this return. Returned by 283 * writepage(); 284 * 285 * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has 286 * unlocked it and the page might have been truncated. 287 * The caller should back up to acquiring a new page and 288 * trying again. The aop will be taking reasonable 289 * precautions not to livelock. If the caller held a page 290 * reference, it should drop it before retrying. Returned 291 * by read_folio(). 292 * 293 * address_space_operation functions return these large constants to indicate 294 * special semantics to the caller. These are much larger than the bytes in a 295 * page to allow for functions that return the number of bytes operated on in a 296 * given page. 297 */ 298 299 enum positive_aop_returns { 300 AOP_WRITEPAGE_ACTIVATE = 0x80000, 301 AOP_TRUNCATED_PAGE = 0x80001, 302 }; 303 304 /* 305 * oh the beauties of C type declarations. 306 */ 307 struct page; 308 struct address_space; 309 struct writeback_control; 310 struct readahead_control; 311 312 /* 313 * Write life time hint values. 314 * Stored in struct inode as u8. 315 */ 316 enum rw_hint { 317 WRITE_LIFE_NOT_SET = 0, 318 WRITE_LIFE_NONE = RWH_WRITE_LIFE_NONE, 319 WRITE_LIFE_SHORT = RWH_WRITE_LIFE_SHORT, 320 WRITE_LIFE_MEDIUM = RWH_WRITE_LIFE_MEDIUM, 321 WRITE_LIFE_LONG = RWH_WRITE_LIFE_LONG, 322 WRITE_LIFE_EXTREME = RWH_WRITE_LIFE_EXTREME, 323 }; 324 325 /* Match RWF_* bits to IOCB bits */ 326 #define IOCB_HIPRI (__force int) RWF_HIPRI 327 #define IOCB_DSYNC (__force int) RWF_DSYNC 328 #define IOCB_SYNC (__force int) RWF_SYNC 329 #define IOCB_NOWAIT (__force int) RWF_NOWAIT 330 #define IOCB_APPEND (__force int) RWF_APPEND 331 332 /* non-RWF related bits - start at 16 */ 333 #define IOCB_EVENTFD (1 << 16) 334 #define IOCB_DIRECT (1 << 17) 335 #define IOCB_WRITE (1 << 18) 336 /* iocb->ki_waitq is valid */ 337 #define IOCB_WAITQ (1 << 19) 338 #define IOCB_NOIO (1 << 20) 339 /* can use bio alloc cache */ 340 #define IOCB_ALLOC_CACHE (1 << 21) 341 342 /* for use in trace events */ 343 #define TRACE_IOCB_STRINGS \ 344 { IOCB_HIPRI, "HIPRI" }, \ 345 { IOCB_DSYNC, "DSYNC" }, \ 346 { IOCB_SYNC, "SYNC" }, \ 347 { IOCB_NOWAIT, "NOWAIT" }, \ 348 { IOCB_APPEND, "APPEND" }, \ 349 { IOCB_EVENTFD, "EVENTFD"}, \ 350 { IOCB_DIRECT, "DIRECT" }, \ 351 { IOCB_WRITE, "WRITE" }, \ 352 { IOCB_WAITQ, "WAITQ" }, \ 353 { IOCB_NOIO, "NOIO" }, \ 354 { IOCB_ALLOC_CACHE, "ALLOC_CACHE" } 355 356 struct kiocb { 357 struct file *ki_filp; 358 loff_t ki_pos; 359 void (*ki_complete)(struct kiocb *iocb, long ret); 360 void *private; 361 int ki_flags; 362 u16 ki_ioprio; /* See linux/ioprio.h */ 363 struct wait_page_queue *ki_waitq; /* for async buffered IO */ 364 }; 365 366 static inline bool is_sync_kiocb(struct kiocb *kiocb) 367 { 368 return kiocb->ki_complete == NULL; 369 } 370 371 struct address_space_operations { 372 int (*writepage)(struct page *page, struct writeback_control *wbc); 373 int (*read_folio)(struct file *, struct folio *); 374 375 /* Write back some dirty pages from this mapping. */ 376 int (*writepages)(struct address_space *, struct writeback_control *); 377 378 /* Mark a folio dirty. Return true if this dirtied it */ 379 bool (*dirty_folio)(struct address_space *, struct folio *); 380 381 void (*readahead)(struct readahead_control *); 382 383 int (*write_begin)(struct file *, struct address_space *mapping, 384 loff_t pos, unsigned len, 385 struct page **pagep, void **fsdata); 386 int (*write_end)(struct file *, struct address_space *mapping, 387 loff_t pos, unsigned len, unsigned copied, 388 struct page *page, void *fsdata); 389 390 /* Unfortunately this kludge is needed for FIBMAP. Don't use it */ 391 sector_t (*bmap)(struct address_space *, sector_t); 392 void (*invalidate_folio) (struct folio *, size_t offset, size_t len); 393 bool (*release_folio)(struct folio *, gfp_t); 394 void (*free_folio)(struct folio *folio); 395 ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter); 396 /* 397 * migrate the contents of a folio to the specified target. If 398 * migrate_mode is MIGRATE_ASYNC, it must not block. 399 */ 400 int (*migrate_folio)(struct address_space *, struct folio *dst, 401 struct folio *src, enum migrate_mode); 402 int (*launder_folio)(struct folio *); 403 bool (*is_partially_uptodate) (struct folio *, size_t from, 404 size_t count); 405 void (*is_dirty_writeback) (struct folio *, bool *dirty, bool *wb); 406 int (*error_remove_page)(struct address_space *, struct page *); 407 408 /* swapfile support */ 409 int (*swap_activate)(struct swap_info_struct *sis, struct file *file, 410 sector_t *span); 411 void (*swap_deactivate)(struct file *file); 412 int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter); 413 }; 414 415 extern const struct address_space_operations empty_aops; 416 417 /** 418 * struct address_space - Contents of a cacheable, mappable object. 419 * @host: Owner, either the inode or the block_device. 420 * @i_pages: Cached pages. 421 * @invalidate_lock: Guards coherency between page cache contents and 422 * file offset->disk block mappings in the filesystem during invalidates. 423 * It is also used to block modification of page cache contents through 424 * memory mappings. 425 * @gfp_mask: Memory allocation flags to use for allocating pages. 426 * @i_mmap_writable: Number of VM_SHARED mappings. 427 * @nr_thps: Number of THPs in the pagecache (non-shmem only). 428 * @i_mmap: Tree of private and shared mappings. 429 * @i_mmap_rwsem: Protects @i_mmap and @i_mmap_writable. 430 * @nrpages: Number of page entries, protected by the i_pages lock. 431 * @writeback_index: Writeback starts here. 432 * @a_ops: Methods. 433 * @flags: Error bits and flags (AS_*). 434 * @wb_err: The most recent error which has occurred. 435 * @private_lock: For use by the owner of the address_space. 436 * @private_list: For use by the owner of the address_space. 437 * @private_data: For use by the owner of the address_space. 438 */ 439 struct address_space { 440 struct inode *host; 441 struct xarray i_pages; 442 struct rw_semaphore invalidate_lock; 443 gfp_t gfp_mask; 444 atomic_t i_mmap_writable; 445 #ifdef CONFIG_READ_ONLY_THP_FOR_FS 446 /* number of thp, only for non-shmem files */ 447 atomic_t nr_thps; 448 #endif 449 struct rb_root_cached i_mmap; 450 struct rw_semaphore i_mmap_rwsem; 451 unsigned long nrpages; 452 pgoff_t writeback_index; 453 const struct address_space_operations *a_ops; 454 unsigned long flags; 455 errseq_t wb_err; 456 spinlock_t private_lock; 457 struct list_head private_list; 458 void *private_data; 459 } __attribute__((aligned(sizeof(long)))) __randomize_layout; 460 /* 461 * On most architectures that alignment is already the case; but 462 * must be enforced here for CRIS, to let the least significant bit 463 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON. 464 */ 465 466 /* XArray tags, for tagging dirty and writeback pages in the pagecache. */ 467 #define PAGECACHE_TAG_DIRTY XA_MARK_0 468 #define PAGECACHE_TAG_WRITEBACK XA_MARK_1 469 #define PAGECACHE_TAG_TOWRITE XA_MARK_2 470 471 /* 472 * Returns true if any of the pages in the mapping are marked with the tag. 473 */ 474 static inline bool mapping_tagged(struct address_space *mapping, xa_mark_t tag) 475 { 476 return xa_marked(&mapping->i_pages, tag); 477 } 478 479 static inline void i_mmap_lock_write(struct address_space *mapping) 480 { 481 down_write(&mapping->i_mmap_rwsem); 482 } 483 484 static inline int i_mmap_trylock_write(struct address_space *mapping) 485 { 486 return down_write_trylock(&mapping->i_mmap_rwsem); 487 } 488 489 static inline void i_mmap_unlock_write(struct address_space *mapping) 490 { 491 up_write(&mapping->i_mmap_rwsem); 492 } 493 494 static inline int i_mmap_trylock_read(struct address_space *mapping) 495 { 496 return down_read_trylock(&mapping->i_mmap_rwsem); 497 } 498 499 static inline void i_mmap_lock_read(struct address_space *mapping) 500 { 501 down_read(&mapping->i_mmap_rwsem); 502 } 503 504 static inline void i_mmap_unlock_read(struct address_space *mapping) 505 { 506 up_read(&mapping->i_mmap_rwsem); 507 } 508 509 static inline void i_mmap_assert_locked(struct address_space *mapping) 510 { 511 lockdep_assert_held(&mapping->i_mmap_rwsem); 512 } 513 514 static inline void i_mmap_assert_write_locked(struct address_space *mapping) 515 { 516 lockdep_assert_held_write(&mapping->i_mmap_rwsem); 517 } 518 519 /* 520 * Might pages of this file be mapped into userspace? 521 */ 522 static inline int mapping_mapped(struct address_space *mapping) 523 { 524 return !RB_EMPTY_ROOT(&mapping->i_mmap.rb_root); 525 } 526 527 /* 528 * Might pages of this file have been modified in userspace? 529 * Note that i_mmap_writable counts all VM_SHARED vmas: do_mmap 530 * marks vma as VM_SHARED if it is shared, and the file was opened for 531 * writing i.e. vma may be mprotected writable even if now readonly. 532 * 533 * If i_mmap_writable is negative, no new writable mappings are allowed. You 534 * can only deny writable mappings, if none exists right now. 535 */ 536 static inline int mapping_writably_mapped(struct address_space *mapping) 537 { 538 return atomic_read(&mapping->i_mmap_writable) > 0; 539 } 540 541 static inline int mapping_map_writable(struct address_space *mapping) 542 { 543 return atomic_inc_unless_negative(&mapping->i_mmap_writable) ? 544 0 : -EPERM; 545 } 546 547 static inline void mapping_unmap_writable(struct address_space *mapping) 548 { 549 atomic_dec(&mapping->i_mmap_writable); 550 } 551 552 static inline int mapping_deny_writable(struct address_space *mapping) 553 { 554 return atomic_dec_unless_positive(&mapping->i_mmap_writable) ? 555 0 : -EBUSY; 556 } 557 558 static inline void mapping_allow_writable(struct address_space *mapping) 559 { 560 atomic_inc(&mapping->i_mmap_writable); 561 } 562 563 /* 564 * Use sequence counter to get consistent i_size on 32-bit processors. 565 */ 566 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 567 #include <linux/seqlock.h> 568 #define __NEED_I_SIZE_ORDERED 569 #define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount) 570 #else 571 #define i_size_ordered_init(inode) do { } while (0) 572 #endif 573 574 struct posix_acl; 575 #define ACL_NOT_CACHED ((void *)(-1)) 576 /* 577 * ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to 578 * cache the ACL. This also means that ->get_inode_acl() can be called in RCU 579 * mode with the LOOKUP_RCU flag. 580 */ 581 #define ACL_DONT_CACHE ((void *)(-3)) 582 583 static inline struct posix_acl * 584 uncached_acl_sentinel(struct task_struct *task) 585 { 586 return (void *)task + 1; 587 } 588 589 static inline bool 590 is_uncached_acl(struct posix_acl *acl) 591 { 592 return (long)acl & 1; 593 } 594 595 #define IOP_FASTPERM 0x0001 596 #define IOP_LOOKUP 0x0002 597 #define IOP_NOFOLLOW 0x0004 598 #define IOP_XATTR 0x0008 599 #define IOP_DEFAULT_READLINK 0x0010 600 601 struct fsnotify_mark_connector; 602 603 /* 604 * Keep mostly read-only and often accessed (especially for 605 * the RCU path lookup and 'stat' data) fields at the beginning 606 * of the 'struct inode' 607 */ 608 struct inode { 609 umode_t i_mode; 610 unsigned short i_opflags; 611 kuid_t i_uid; 612 kgid_t i_gid; 613 unsigned int i_flags; 614 615 #ifdef CONFIG_FS_POSIX_ACL 616 struct posix_acl *i_acl; 617 struct posix_acl *i_default_acl; 618 #endif 619 620 const struct inode_operations *i_op; 621 struct super_block *i_sb; 622 struct address_space *i_mapping; 623 624 #ifdef CONFIG_SECURITY 625 void *i_security; 626 #endif 627 628 /* Stat data, not accessed from path walking */ 629 unsigned long i_ino; 630 /* 631 * Filesystems may only read i_nlink directly. They shall use the 632 * following functions for modification: 633 * 634 * (set|clear|inc|drop)_nlink 635 * inode_(inc|dec)_link_count 636 */ 637 union { 638 const unsigned int i_nlink; 639 unsigned int __i_nlink; 640 }; 641 dev_t i_rdev; 642 loff_t i_size; 643 struct timespec64 i_atime; 644 struct timespec64 i_mtime; 645 struct timespec64 __i_ctime; /* use inode_*_ctime accessors! */ 646 spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */ 647 unsigned short i_bytes; 648 u8 i_blkbits; 649 u8 i_write_hint; 650 blkcnt_t i_blocks; 651 652 #ifdef __NEED_I_SIZE_ORDERED 653 seqcount_t i_size_seqcount; 654 #endif 655 656 /* Misc */ 657 unsigned long i_state; 658 struct rw_semaphore i_rwsem; 659 660 unsigned long dirtied_when; /* jiffies of first dirtying */ 661 unsigned long dirtied_time_when; 662 663 struct hlist_node i_hash; 664 struct list_head i_io_list; /* backing dev IO list */ 665 #ifdef CONFIG_CGROUP_WRITEBACK 666 struct bdi_writeback *i_wb; /* the associated cgroup wb */ 667 668 /* foreign inode detection, see wbc_detach_inode() */ 669 int i_wb_frn_winner; 670 u16 i_wb_frn_avg_time; 671 u16 i_wb_frn_history; 672 #endif 673 struct list_head i_lru; /* inode LRU list */ 674 struct list_head i_sb_list; 675 struct list_head i_wb_list; /* backing dev writeback list */ 676 union { 677 struct hlist_head i_dentry; 678 struct rcu_head i_rcu; 679 }; 680 atomic64_t i_version; 681 atomic64_t i_sequence; /* see futex */ 682 atomic_t i_count; 683 atomic_t i_dio_count; 684 atomic_t i_writecount; 685 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING) 686 atomic_t i_readcount; /* struct files open RO */ 687 #endif 688 union { 689 const struct file_operations *i_fop; /* former ->i_op->default_file_ops */ 690 void (*free_inode)(struct inode *); 691 }; 692 struct file_lock_context *i_flctx; 693 struct address_space i_data; 694 struct list_head i_devices; 695 union { 696 struct pipe_inode_info *i_pipe; 697 struct cdev *i_cdev; 698 char *i_link; 699 unsigned i_dir_seq; 700 }; 701 702 __u32 i_generation; 703 704 #ifdef CONFIG_FSNOTIFY 705 __u32 i_fsnotify_mask; /* all events this inode cares about */ 706 struct fsnotify_mark_connector __rcu *i_fsnotify_marks; 707 #endif 708 709 #ifdef CONFIG_FS_ENCRYPTION 710 struct fscrypt_info *i_crypt_info; 711 #endif 712 713 #ifdef CONFIG_FS_VERITY 714 struct fsverity_info *i_verity_info; 715 #endif 716 717 void *i_private; /* fs or device private pointer */ 718 } __randomize_layout; 719 720 struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode); 721 722 static inline unsigned int i_blocksize(const struct inode *node) 723 { 724 return (1 << node->i_blkbits); 725 } 726 727 static inline int inode_unhashed(struct inode *inode) 728 { 729 return hlist_unhashed(&inode->i_hash); 730 } 731 732 /* 733 * __mark_inode_dirty expects inodes to be hashed. Since we don't 734 * want special inodes in the fileset inode space, we make them 735 * appear hashed, but do not put on any lists. hlist_del() 736 * will work fine and require no locking. 737 */ 738 static inline void inode_fake_hash(struct inode *inode) 739 { 740 hlist_add_fake(&inode->i_hash); 741 } 742 743 /* 744 * inode->i_mutex nesting subclasses for the lock validator: 745 * 746 * 0: the object of the current VFS operation 747 * 1: parent 748 * 2: child/target 749 * 3: xattr 750 * 4: second non-directory 751 * 5: second parent (when locking independent directories in rename) 752 * 753 * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two 754 * non-directories at once. 755 * 756 * The locking order between these classes is 757 * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory 758 */ 759 enum inode_i_mutex_lock_class 760 { 761 I_MUTEX_NORMAL, 762 I_MUTEX_PARENT, 763 I_MUTEX_CHILD, 764 I_MUTEX_XATTR, 765 I_MUTEX_NONDIR2, 766 I_MUTEX_PARENT2, 767 }; 768 769 static inline void inode_lock(struct inode *inode) 770 { 771 down_write(&inode->i_rwsem); 772 } 773 774 static inline void inode_unlock(struct inode *inode) 775 { 776 up_write(&inode->i_rwsem); 777 } 778 779 static inline void inode_lock_shared(struct inode *inode) 780 { 781 down_read(&inode->i_rwsem); 782 } 783 784 static inline void inode_unlock_shared(struct inode *inode) 785 { 786 up_read(&inode->i_rwsem); 787 } 788 789 static inline int inode_trylock(struct inode *inode) 790 { 791 return down_write_trylock(&inode->i_rwsem); 792 } 793 794 static inline int inode_trylock_shared(struct inode *inode) 795 { 796 return down_read_trylock(&inode->i_rwsem); 797 } 798 799 static inline int inode_is_locked(struct inode *inode) 800 { 801 return rwsem_is_locked(&inode->i_rwsem); 802 } 803 804 static inline void inode_lock_nested(struct inode *inode, unsigned subclass) 805 { 806 down_write_nested(&inode->i_rwsem, subclass); 807 } 808 809 static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass) 810 { 811 down_read_nested(&inode->i_rwsem, subclass); 812 } 813 814 static inline void filemap_invalidate_lock(struct address_space *mapping) 815 { 816 down_write(&mapping->invalidate_lock); 817 } 818 819 static inline void filemap_invalidate_unlock(struct address_space *mapping) 820 { 821 up_write(&mapping->invalidate_lock); 822 } 823 824 static inline void filemap_invalidate_lock_shared(struct address_space *mapping) 825 { 826 down_read(&mapping->invalidate_lock); 827 } 828 829 static inline int filemap_invalidate_trylock_shared( 830 struct address_space *mapping) 831 { 832 return down_read_trylock(&mapping->invalidate_lock); 833 } 834 835 static inline void filemap_invalidate_unlock_shared( 836 struct address_space *mapping) 837 { 838 up_read(&mapping->invalidate_lock); 839 } 840 841 void lock_two_nondirectories(struct inode *, struct inode*); 842 void unlock_two_nondirectories(struct inode *, struct inode*); 843 844 void filemap_invalidate_lock_two(struct address_space *mapping1, 845 struct address_space *mapping2); 846 void filemap_invalidate_unlock_two(struct address_space *mapping1, 847 struct address_space *mapping2); 848 849 850 /* 851 * NOTE: in a 32bit arch with a preemptable kernel and 852 * an UP compile the i_size_read/write must be atomic 853 * with respect to the local cpu (unlike with preempt disabled), 854 * but they don't need to be atomic with respect to other cpus like in 855 * true SMP (so they need either to either locally disable irq around 856 * the read or for example on x86 they can be still implemented as a 857 * cmpxchg8b without the need of the lock prefix). For SMP compiles 858 * and 64bit archs it makes no difference if preempt is enabled or not. 859 */ 860 static inline loff_t i_size_read(const struct inode *inode) 861 { 862 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 863 loff_t i_size; 864 unsigned int seq; 865 866 do { 867 seq = read_seqcount_begin(&inode->i_size_seqcount); 868 i_size = inode->i_size; 869 } while (read_seqcount_retry(&inode->i_size_seqcount, seq)); 870 return i_size; 871 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION) 872 loff_t i_size; 873 874 preempt_disable(); 875 i_size = inode->i_size; 876 preempt_enable(); 877 return i_size; 878 #else 879 return inode->i_size; 880 #endif 881 } 882 883 /* 884 * NOTE: unlike i_size_read(), i_size_write() does need locking around it 885 * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount 886 * can be lost, resulting in subsequent i_size_read() calls spinning forever. 887 */ 888 static inline void i_size_write(struct inode *inode, loff_t i_size) 889 { 890 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 891 preempt_disable(); 892 write_seqcount_begin(&inode->i_size_seqcount); 893 inode->i_size = i_size; 894 write_seqcount_end(&inode->i_size_seqcount); 895 preempt_enable(); 896 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION) 897 preempt_disable(); 898 inode->i_size = i_size; 899 preempt_enable(); 900 #else 901 inode->i_size = i_size; 902 #endif 903 } 904 905 static inline unsigned iminor(const struct inode *inode) 906 { 907 return MINOR(inode->i_rdev); 908 } 909 910 static inline unsigned imajor(const struct inode *inode) 911 { 912 return MAJOR(inode->i_rdev); 913 } 914 915 struct fown_struct { 916 rwlock_t lock; /* protects pid, uid, euid fields */ 917 struct pid *pid; /* pid or -pgrp where SIGIO should be sent */ 918 enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */ 919 kuid_t uid, euid; /* uid/euid of process setting the owner */ 920 int signum; /* posix.1b rt signal to be delivered on IO */ 921 }; 922 923 /** 924 * struct file_ra_state - Track a file's readahead state. 925 * @start: Where the most recent readahead started. 926 * @size: Number of pages read in the most recent readahead. 927 * @async_size: Numer of pages that were/are not needed immediately 928 * and so were/are genuinely "ahead". Start next readahead when 929 * the first of these pages is accessed. 930 * @ra_pages: Maximum size of a readahead request, copied from the bdi. 931 * @mmap_miss: How many mmap accesses missed in the page cache. 932 * @prev_pos: The last byte in the most recent read request. 933 * 934 * When this structure is passed to ->readahead(), the "most recent" 935 * readahead means the current readahead. 936 */ 937 struct file_ra_state { 938 pgoff_t start; 939 unsigned int size; 940 unsigned int async_size; 941 unsigned int ra_pages; 942 unsigned int mmap_miss; 943 loff_t prev_pos; 944 }; 945 946 /* 947 * Check if @index falls in the readahead windows. 948 */ 949 static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index) 950 { 951 return (index >= ra->start && 952 index < ra->start + ra->size); 953 } 954 955 /* 956 * f_{lock,count,pos_lock} members can be highly contended and share 957 * the same cacheline. f_{lock,mode} are very frequently used together 958 * and so share the same cacheline as well. The read-mostly 959 * f_{path,inode,op} are kept on a separate cacheline. 960 */ 961 struct file { 962 union { 963 struct llist_node f_llist; 964 struct rcu_head f_rcuhead; 965 unsigned int f_iocb_flags; 966 }; 967 968 /* 969 * Protects f_ep, f_flags. 970 * Must not be taken from IRQ context. 971 */ 972 spinlock_t f_lock; 973 fmode_t f_mode; 974 atomic_long_t f_count; 975 struct mutex f_pos_lock; 976 loff_t f_pos; 977 unsigned int f_flags; 978 struct fown_struct f_owner; 979 const struct cred *f_cred; 980 struct file_ra_state f_ra; 981 struct path f_path; 982 struct inode *f_inode; /* cached value */ 983 const struct file_operations *f_op; 984 985 u64 f_version; 986 #ifdef CONFIG_SECURITY 987 void *f_security; 988 #endif 989 /* needed for tty driver, and maybe others */ 990 void *private_data; 991 992 #ifdef CONFIG_EPOLL 993 /* Used by fs/eventpoll.c to link all the hooks to this file */ 994 struct hlist_head *f_ep; 995 #endif /* #ifdef CONFIG_EPOLL */ 996 struct address_space *f_mapping; 997 errseq_t f_wb_err; 998 errseq_t f_sb_err; /* for syncfs */ 999 } __randomize_layout 1000 __attribute__((aligned(4))); /* lest something weird decides that 2 is OK */ 1001 1002 struct file_handle { 1003 __u32 handle_bytes; 1004 int handle_type; 1005 /* file identifier */ 1006 unsigned char f_handle[]; 1007 }; 1008 1009 static inline struct file *get_file(struct file *f) 1010 { 1011 atomic_long_inc(&f->f_count); 1012 return f; 1013 } 1014 #define get_file_rcu(x) atomic_long_inc_not_zero(&(x)->f_count) 1015 #define file_count(x) atomic_long_read(&(x)->f_count) 1016 1017 #define MAX_NON_LFS ((1UL<<31) - 1) 1018 1019 /* Page cache limit. The filesystems should put that into their s_maxbytes 1020 limits, otherwise bad things can happen in VM. */ 1021 #if BITS_PER_LONG==32 1022 #define MAX_LFS_FILESIZE ((loff_t)ULONG_MAX << PAGE_SHIFT) 1023 #elif BITS_PER_LONG==64 1024 #define MAX_LFS_FILESIZE ((loff_t)LLONG_MAX) 1025 #endif 1026 1027 /* legacy typedef, should eventually be removed */ 1028 typedef void *fl_owner_t; 1029 1030 struct file_lock; 1031 1032 /* The following constant reflects the upper bound of the file/locking space */ 1033 #ifndef OFFSET_MAX 1034 #define OFFSET_MAX type_max(loff_t) 1035 #define OFFT_OFFSET_MAX type_max(off_t) 1036 #endif 1037 1038 extern void send_sigio(struct fown_struct *fown, int fd, int band); 1039 1040 static inline struct inode *file_inode(const struct file *f) 1041 { 1042 return f->f_inode; 1043 } 1044 1045 static inline struct dentry *file_dentry(const struct file *file) 1046 { 1047 return d_real(file->f_path.dentry, file_inode(file)); 1048 } 1049 1050 struct fasync_struct { 1051 rwlock_t fa_lock; 1052 int magic; 1053 int fa_fd; 1054 struct fasync_struct *fa_next; /* singly linked list */ 1055 struct file *fa_file; 1056 struct rcu_head fa_rcu; 1057 }; 1058 1059 #define FASYNC_MAGIC 0x4601 1060 1061 /* SMP safe fasync helpers: */ 1062 extern int fasync_helper(int, struct file *, int, struct fasync_struct **); 1063 extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *); 1064 extern int fasync_remove_entry(struct file *, struct fasync_struct **); 1065 extern struct fasync_struct *fasync_alloc(void); 1066 extern void fasync_free(struct fasync_struct *); 1067 1068 /* can be called from interrupts */ 1069 extern void kill_fasync(struct fasync_struct **, int, int); 1070 1071 extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force); 1072 extern int f_setown(struct file *filp, unsigned long arg, int force); 1073 extern void f_delown(struct file *filp); 1074 extern pid_t f_getown(struct file *filp); 1075 extern int send_sigurg(struct fown_struct *fown); 1076 1077 /* 1078 * sb->s_flags. Note that these mirror the equivalent MS_* flags where 1079 * represented in both. 1080 */ 1081 #define SB_RDONLY BIT(0) /* Mount read-only */ 1082 #define SB_NOSUID BIT(1) /* Ignore suid and sgid bits */ 1083 #define SB_NODEV BIT(2) /* Disallow access to device special files */ 1084 #define SB_NOEXEC BIT(3) /* Disallow program execution */ 1085 #define SB_SYNCHRONOUS BIT(4) /* Writes are synced at once */ 1086 #define SB_MANDLOCK BIT(6) /* Allow mandatory locks on an FS */ 1087 #define SB_DIRSYNC BIT(7) /* Directory modifications are synchronous */ 1088 #define SB_NOATIME BIT(10) /* Do not update access times. */ 1089 #define SB_NODIRATIME BIT(11) /* Do not update directory access times */ 1090 #define SB_SILENT BIT(15) 1091 #define SB_POSIXACL BIT(16) /* VFS does not apply the umask */ 1092 #define SB_INLINECRYPT BIT(17) /* Use blk-crypto for encrypted files */ 1093 #define SB_KERNMOUNT BIT(22) /* this is a kern_mount call */ 1094 #define SB_I_VERSION BIT(23) /* Update inode I_version field */ 1095 #define SB_LAZYTIME BIT(25) /* Update the on-disk [acm]times lazily */ 1096 1097 /* These sb flags are internal to the kernel */ 1098 #define SB_SUBMOUNT BIT(26) 1099 #define SB_FORCE BIT(27) 1100 #define SB_NOSEC BIT(28) 1101 #define SB_BORN BIT(29) 1102 #define SB_ACTIVE BIT(30) 1103 #define SB_NOUSER BIT(31) 1104 1105 /* These flags relate to encoding and casefolding */ 1106 #define SB_ENC_STRICT_MODE_FL (1 << 0) 1107 1108 #define sb_has_strict_encoding(sb) \ 1109 (sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL) 1110 1111 /* 1112 * Umount options 1113 */ 1114 1115 #define MNT_FORCE 0x00000001 /* Attempt to forcibily umount */ 1116 #define MNT_DETACH 0x00000002 /* Just detach from the tree */ 1117 #define MNT_EXPIRE 0x00000004 /* Mark for expiry */ 1118 #define UMOUNT_NOFOLLOW 0x00000008 /* Don't follow symlink on umount */ 1119 #define UMOUNT_UNUSED 0x80000000 /* Flag guaranteed to be unused */ 1120 1121 /* sb->s_iflags */ 1122 #define SB_I_CGROUPWB 0x00000001 /* cgroup-aware writeback enabled */ 1123 #define SB_I_NOEXEC 0x00000002 /* Ignore executables on this fs */ 1124 #define SB_I_NODEV 0x00000004 /* Ignore devices on this fs */ 1125 #define SB_I_STABLE_WRITES 0x00000008 /* don't modify blks until WB is done */ 1126 1127 /* sb->s_iflags to limit user namespace mounts */ 1128 #define SB_I_USERNS_VISIBLE 0x00000010 /* fstype already mounted */ 1129 #define SB_I_IMA_UNVERIFIABLE_SIGNATURE 0x00000020 1130 #define SB_I_UNTRUSTED_MOUNTER 0x00000040 1131 1132 #define SB_I_SKIP_SYNC 0x00000100 /* Skip superblock at global sync */ 1133 #define SB_I_PERSB_BDI 0x00000200 /* has a per-sb bdi */ 1134 #define SB_I_TS_EXPIRY_WARNED 0x00000400 /* warned about timestamp range expiry */ 1135 #define SB_I_RETIRED 0x00000800 /* superblock shouldn't be reused */ 1136 1137 /* Possible states of 'frozen' field */ 1138 enum { 1139 SB_UNFROZEN = 0, /* FS is unfrozen */ 1140 SB_FREEZE_WRITE = 1, /* Writes, dir ops, ioctls frozen */ 1141 SB_FREEZE_PAGEFAULT = 2, /* Page faults stopped as well */ 1142 SB_FREEZE_FS = 3, /* For internal FS use (e.g. to stop 1143 * internal threads if needed) */ 1144 SB_FREEZE_COMPLETE = 4, /* ->freeze_fs finished successfully */ 1145 }; 1146 1147 #define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1) 1148 1149 struct sb_writers { 1150 int frozen; /* Is sb frozen? */ 1151 struct percpu_rw_semaphore rw_sem[SB_FREEZE_LEVELS]; 1152 }; 1153 1154 struct super_block { 1155 struct list_head s_list; /* Keep this first */ 1156 dev_t s_dev; /* search index; _not_ kdev_t */ 1157 unsigned char s_blocksize_bits; 1158 unsigned long s_blocksize; 1159 loff_t s_maxbytes; /* Max file size */ 1160 struct file_system_type *s_type; 1161 const struct super_operations *s_op; 1162 const struct dquot_operations *dq_op; 1163 const struct quotactl_ops *s_qcop; 1164 const struct export_operations *s_export_op; 1165 unsigned long s_flags; 1166 unsigned long s_iflags; /* internal SB_I_* flags */ 1167 unsigned long s_magic; 1168 struct dentry *s_root; 1169 struct rw_semaphore s_umount; 1170 int s_count; 1171 atomic_t s_active; 1172 #ifdef CONFIG_SECURITY 1173 void *s_security; 1174 #endif 1175 const struct xattr_handler **s_xattr; 1176 #ifdef CONFIG_FS_ENCRYPTION 1177 const struct fscrypt_operations *s_cop; 1178 struct fscrypt_keyring *s_master_keys; /* master crypto keys in use */ 1179 #endif 1180 #ifdef CONFIG_FS_VERITY 1181 const struct fsverity_operations *s_vop; 1182 #endif 1183 #if IS_ENABLED(CONFIG_UNICODE) 1184 struct unicode_map *s_encoding; 1185 __u16 s_encoding_flags; 1186 #endif 1187 struct hlist_bl_head s_roots; /* alternate root dentries for NFS */ 1188 struct list_head s_mounts; /* list of mounts; _not_ for fs use */ 1189 struct block_device *s_bdev; 1190 struct backing_dev_info *s_bdi; 1191 struct mtd_info *s_mtd; 1192 struct hlist_node s_instances; 1193 unsigned int s_quota_types; /* Bitmask of supported quota types */ 1194 struct quota_info s_dquot; /* Diskquota specific options */ 1195 1196 struct sb_writers s_writers; 1197 1198 /* 1199 * Keep s_fs_info, s_time_gran, s_fsnotify_mask, and 1200 * s_fsnotify_marks together for cache efficiency. They are frequently 1201 * accessed and rarely modified. 1202 */ 1203 void *s_fs_info; /* Filesystem private info */ 1204 1205 /* Granularity of c/m/atime in ns (cannot be worse than a second) */ 1206 u32 s_time_gran; 1207 /* Time limits for c/m/atime in seconds */ 1208 time64_t s_time_min; 1209 time64_t s_time_max; 1210 #ifdef CONFIG_FSNOTIFY 1211 __u32 s_fsnotify_mask; 1212 struct fsnotify_mark_connector __rcu *s_fsnotify_marks; 1213 #endif 1214 1215 char s_id[32]; /* Informational name */ 1216 uuid_t s_uuid; /* UUID */ 1217 1218 unsigned int s_max_links; 1219 1220 /* 1221 * The next field is for VFS *only*. No filesystems have any business 1222 * even looking at it. You had been warned. 1223 */ 1224 struct mutex s_vfs_rename_mutex; /* Kludge */ 1225 1226 /* 1227 * Filesystem subtype. If non-empty the filesystem type field 1228 * in /proc/mounts will be "type.subtype" 1229 */ 1230 const char *s_subtype; 1231 1232 const struct dentry_operations *s_d_op; /* default d_op for dentries */ 1233 1234 struct shrinker s_shrink; /* per-sb shrinker handle */ 1235 1236 /* Number of inodes with nlink == 0 but still referenced */ 1237 atomic_long_t s_remove_count; 1238 1239 /* 1240 * Number of inode/mount/sb objects that are being watched, note that 1241 * inodes objects are currently double-accounted. 1242 */ 1243 atomic_long_t s_fsnotify_connectors; 1244 1245 /* Read-only state of the superblock is being changed */ 1246 int s_readonly_remount; 1247 1248 /* per-sb errseq_t for reporting writeback errors via syncfs */ 1249 errseq_t s_wb_err; 1250 1251 /* AIO completions deferred from interrupt context */ 1252 struct workqueue_struct *s_dio_done_wq; 1253 struct hlist_head s_pins; 1254 1255 /* 1256 * Owning user namespace and default context in which to 1257 * interpret filesystem uids, gids, quotas, device nodes, 1258 * xattrs and security labels. 1259 */ 1260 struct user_namespace *s_user_ns; 1261 1262 /* 1263 * The list_lru structure is essentially just a pointer to a table 1264 * of per-node lru lists, each of which has its own spinlock. 1265 * There is no need to put them into separate cachelines. 1266 */ 1267 struct list_lru s_dentry_lru; 1268 struct list_lru s_inode_lru; 1269 struct rcu_head rcu; 1270 struct work_struct destroy_work; 1271 1272 struct mutex s_sync_lock; /* sync serialisation lock */ 1273 1274 /* 1275 * Indicates how deep in a filesystem stack this SB is 1276 */ 1277 int s_stack_depth; 1278 1279 /* s_inode_list_lock protects s_inodes */ 1280 spinlock_t s_inode_list_lock ____cacheline_aligned_in_smp; 1281 struct list_head s_inodes; /* all inodes */ 1282 1283 spinlock_t s_inode_wblist_lock; 1284 struct list_head s_inodes_wb; /* writeback inodes */ 1285 } __randomize_layout; 1286 1287 static inline struct user_namespace *i_user_ns(const struct inode *inode) 1288 { 1289 return inode->i_sb->s_user_ns; 1290 } 1291 1292 /* Helper functions so that in most cases filesystems will 1293 * not need to deal directly with kuid_t and kgid_t and can 1294 * instead deal with the raw numeric values that are stored 1295 * in the filesystem. 1296 */ 1297 static inline uid_t i_uid_read(const struct inode *inode) 1298 { 1299 return from_kuid(i_user_ns(inode), inode->i_uid); 1300 } 1301 1302 static inline gid_t i_gid_read(const struct inode *inode) 1303 { 1304 return from_kgid(i_user_ns(inode), inode->i_gid); 1305 } 1306 1307 static inline void i_uid_write(struct inode *inode, uid_t uid) 1308 { 1309 inode->i_uid = make_kuid(i_user_ns(inode), uid); 1310 } 1311 1312 static inline void i_gid_write(struct inode *inode, gid_t gid) 1313 { 1314 inode->i_gid = make_kgid(i_user_ns(inode), gid); 1315 } 1316 1317 /** 1318 * i_uid_into_vfsuid - map an inode's i_uid down according to an idmapping 1319 * @idmap: idmap of the mount the inode was found from 1320 * @inode: inode to map 1321 * 1322 * Return: whe inode's i_uid mapped down according to @idmap. 1323 * If the inode's i_uid has no mapping INVALID_VFSUID is returned. 1324 */ 1325 static inline vfsuid_t i_uid_into_vfsuid(struct mnt_idmap *idmap, 1326 const struct inode *inode) 1327 { 1328 return make_vfsuid(idmap, i_user_ns(inode), inode->i_uid); 1329 } 1330 1331 /** 1332 * i_uid_needs_update - check whether inode's i_uid needs to be updated 1333 * @idmap: idmap of the mount the inode was found from 1334 * @attr: the new attributes of @inode 1335 * @inode: the inode to update 1336 * 1337 * Check whether the $inode's i_uid field needs to be updated taking idmapped 1338 * mounts into account if the filesystem supports it. 1339 * 1340 * Return: true if @inode's i_uid field needs to be updated, false if not. 1341 */ 1342 static inline bool i_uid_needs_update(struct mnt_idmap *idmap, 1343 const struct iattr *attr, 1344 const struct inode *inode) 1345 { 1346 return ((attr->ia_valid & ATTR_UID) && 1347 !vfsuid_eq(attr->ia_vfsuid, 1348 i_uid_into_vfsuid(idmap, inode))); 1349 } 1350 1351 /** 1352 * i_uid_update - update @inode's i_uid field 1353 * @idmap: idmap of the mount the inode was found from 1354 * @attr: the new attributes of @inode 1355 * @inode: the inode to update 1356 * 1357 * Safely update @inode's i_uid field translating the vfsuid of any idmapped 1358 * mount into the filesystem kuid. 1359 */ 1360 static inline void i_uid_update(struct mnt_idmap *idmap, 1361 const struct iattr *attr, 1362 struct inode *inode) 1363 { 1364 if (attr->ia_valid & ATTR_UID) 1365 inode->i_uid = from_vfsuid(idmap, i_user_ns(inode), 1366 attr->ia_vfsuid); 1367 } 1368 1369 /** 1370 * i_gid_into_vfsgid - map an inode's i_gid down according to an idmapping 1371 * @idmap: idmap of the mount the inode was found from 1372 * @inode: inode to map 1373 * 1374 * Return: the inode's i_gid mapped down according to @idmap. 1375 * If the inode's i_gid has no mapping INVALID_VFSGID is returned. 1376 */ 1377 static inline vfsgid_t i_gid_into_vfsgid(struct mnt_idmap *idmap, 1378 const struct inode *inode) 1379 { 1380 return make_vfsgid(idmap, i_user_ns(inode), inode->i_gid); 1381 } 1382 1383 /** 1384 * i_gid_needs_update - check whether inode's i_gid needs to be updated 1385 * @idmap: idmap of the mount the inode was found from 1386 * @attr: the new attributes of @inode 1387 * @inode: the inode to update 1388 * 1389 * Check whether the $inode's i_gid field needs to be updated taking idmapped 1390 * mounts into account if the filesystem supports it. 1391 * 1392 * Return: true if @inode's i_gid field needs to be updated, false if not. 1393 */ 1394 static inline bool i_gid_needs_update(struct mnt_idmap *idmap, 1395 const struct iattr *attr, 1396 const struct inode *inode) 1397 { 1398 return ((attr->ia_valid & ATTR_GID) && 1399 !vfsgid_eq(attr->ia_vfsgid, 1400 i_gid_into_vfsgid(idmap, inode))); 1401 } 1402 1403 /** 1404 * i_gid_update - update @inode's i_gid field 1405 * @idmap: idmap of the mount the inode was found from 1406 * @attr: the new attributes of @inode 1407 * @inode: the inode to update 1408 * 1409 * Safely update @inode's i_gid field translating the vfsgid of any idmapped 1410 * mount into the filesystem kgid. 1411 */ 1412 static inline void i_gid_update(struct mnt_idmap *idmap, 1413 const struct iattr *attr, 1414 struct inode *inode) 1415 { 1416 if (attr->ia_valid & ATTR_GID) 1417 inode->i_gid = from_vfsgid(idmap, i_user_ns(inode), 1418 attr->ia_vfsgid); 1419 } 1420 1421 /** 1422 * inode_fsuid_set - initialize inode's i_uid field with callers fsuid 1423 * @inode: inode to initialize 1424 * @idmap: idmap of the mount the inode was found from 1425 * 1426 * Initialize the i_uid field of @inode. If the inode was found/created via 1427 * an idmapped mount map the caller's fsuid according to @idmap. 1428 */ 1429 static inline void inode_fsuid_set(struct inode *inode, 1430 struct mnt_idmap *idmap) 1431 { 1432 inode->i_uid = mapped_fsuid(idmap, i_user_ns(inode)); 1433 } 1434 1435 /** 1436 * inode_fsgid_set - initialize inode's i_gid field with callers fsgid 1437 * @inode: inode to initialize 1438 * @idmap: idmap of the mount the inode was found from 1439 * 1440 * Initialize the i_gid field of @inode. If the inode was found/created via 1441 * an idmapped mount map the caller's fsgid according to @idmap. 1442 */ 1443 static inline void inode_fsgid_set(struct inode *inode, 1444 struct mnt_idmap *idmap) 1445 { 1446 inode->i_gid = mapped_fsgid(idmap, i_user_ns(inode)); 1447 } 1448 1449 /** 1450 * fsuidgid_has_mapping() - check whether caller's fsuid/fsgid is mapped 1451 * @sb: the superblock we want a mapping in 1452 * @idmap: idmap of the relevant mount 1453 * 1454 * Check whether the caller's fsuid and fsgid have a valid mapping in the 1455 * s_user_ns of the superblock @sb. If the caller is on an idmapped mount map 1456 * the caller's fsuid and fsgid according to the @idmap first. 1457 * 1458 * Return: true if fsuid and fsgid is mapped, false if not. 1459 */ 1460 static inline bool fsuidgid_has_mapping(struct super_block *sb, 1461 struct mnt_idmap *idmap) 1462 { 1463 struct user_namespace *fs_userns = sb->s_user_ns; 1464 kuid_t kuid; 1465 kgid_t kgid; 1466 1467 kuid = mapped_fsuid(idmap, fs_userns); 1468 if (!uid_valid(kuid)) 1469 return false; 1470 kgid = mapped_fsgid(idmap, fs_userns); 1471 if (!gid_valid(kgid)) 1472 return false; 1473 return kuid_has_mapping(fs_userns, kuid) && 1474 kgid_has_mapping(fs_userns, kgid); 1475 } 1476 1477 struct timespec64 current_mgtime(struct inode *inode); 1478 struct timespec64 current_time(struct inode *inode); 1479 struct timespec64 inode_set_ctime_current(struct inode *inode); 1480 1481 /* 1482 * Multigrain timestamps 1483 * 1484 * Conditionally use fine-grained ctime and mtime timestamps when there 1485 * are users actively observing them via getattr. The primary use-case 1486 * for this is NFS clients that use the ctime to distinguish between 1487 * different states of the file, and that are often fooled by multiple 1488 * operations that occur in the same coarse-grained timer tick. 1489 * 1490 * The kernel always keeps normalized struct timespec64 values in the ctime, 1491 * which means that only the first 30 bits of the value are used. Use the 1492 * 31st bit of the ctime's tv_nsec field as a flag to indicate that the value 1493 * has been queried since it was last updated. 1494 */ 1495 #define I_CTIME_QUERIED (1L<<30) 1496 1497 /** 1498 * inode_get_ctime - fetch the current ctime from the inode 1499 * @inode: inode from which to fetch ctime 1500 * 1501 * Grab the current ctime tv_nsec field from the inode, mask off the 1502 * I_CTIME_QUERIED flag and return it. This is mostly intended for use by 1503 * internal consumers of the ctime that aren't concerned with ensuring a 1504 * fine-grained update on the next change (e.g. when preparing to store 1505 * the value in the backing store for later retrieval). 1506 * 1507 * This is safe to call regardless of whether the underlying filesystem 1508 * is using multigrain timestamps. 1509 */ 1510 static inline struct timespec64 inode_get_ctime(const struct inode *inode) 1511 { 1512 struct timespec64 ctime; 1513 1514 ctime.tv_sec = inode->__i_ctime.tv_sec; 1515 ctime.tv_nsec = inode->__i_ctime.tv_nsec & ~I_CTIME_QUERIED; 1516 1517 return ctime; 1518 } 1519 1520 /** 1521 * inode_set_ctime_to_ts - set the ctime in the inode 1522 * @inode: inode in which to set the ctime 1523 * @ts: value to set in the ctime field 1524 * 1525 * Set the ctime in @inode to @ts 1526 */ 1527 static inline struct timespec64 inode_set_ctime_to_ts(struct inode *inode, 1528 struct timespec64 ts) 1529 { 1530 inode->__i_ctime = ts; 1531 return ts; 1532 } 1533 1534 /** 1535 * inode_set_ctime - set the ctime in the inode 1536 * @inode: inode in which to set the ctime 1537 * @sec: tv_sec value to set 1538 * @nsec: tv_nsec value to set 1539 * 1540 * Set the ctime in @inode to { @sec, @nsec } 1541 */ 1542 static inline struct timespec64 inode_set_ctime(struct inode *inode, 1543 time64_t sec, long nsec) 1544 { 1545 struct timespec64 ts = { .tv_sec = sec, 1546 .tv_nsec = nsec }; 1547 1548 return inode_set_ctime_to_ts(inode, ts); 1549 } 1550 1551 /* 1552 * Snapshotting support. 1553 */ 1554 1555 /* 1556 * These are internal functions, please use sb_start_{write,pagefault,intwrite} 1557 * instead. 1558 */ 1559 static inline void __sb_end_write(struct super_block *sb, int level) 1560 { 1561 percpu_up_read(sb->s_writers.rw_sem + level-1); 1562 } 1563 1564 static inline void __sb_start_write(struct super_block *sb, int level) 1565 { 1566 percpu_down_read(sb->s_writers.rw_sem + level - 1); 1567 } 1568 1569 static inline bool __sb_start_write_trylock(struct super_block *sb, int level) 1570 { 1571 return percpu_down_read_trylock(sb->s_writers.rw_sem + level - 1); 1572 } 1573 1574 #define __sb_writers_acquired(sb, lev) \ 1575 percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_) 1576 #define __sb_writers_release(sb, lev) \ 1577 percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_) 1578 1579 static inline bool sb_write_started(const struct super_block *sb) 1580 { 1581 return lockdep_is_held_type(sb->s_writers.rw_sem + SB_FREEZE_WRITE - 1, 1); 1582 } 1583 1584 /** 1585 * sb_end_write - drop write access to a superblock 1586 * @sb: the super we wrote to 1587 * 1588 * Decrement number of writers to the filesystem. Wake up possible waiters 1589 * wanting to freeze the filesystem. 1590 */ 1591 static inline void sb_end_write(struct super_block *sb) 1592 { 1593 __sb_end_write(sb, SB_FREEZE_WRITE); 1594 } 1595 1596 /** 1597 * sb_end_pagefault - drop write access to a superblock from a page fault 1598 * @sb: the super we wrote to 1599 * 1600 * Decrement number of processes handling write page fault to the filesystem. 1601 * Wake up possible waiters wanting to freeze the filesystem. 1602 */ 1603 static inline void sb_end_pagefault(struct super_block *sb) 1604 { 1605 __sb_end_write(sb, SB_FREEZE_PAGEFAULT); 1606 } 1607 1608 /** 1609 * sb_end_intwrite - drop write access to a superblock for internal fs purposes 1610 * @sb: the super we wrote to 1611 * 1612 * Decrement fs-internal number of writers to the filesystem. Wake up possible 1613 * waiters wanting to freeze the filesystem. 1614 */ 1615 static inline void sb_end_intwrite(struct super_block *sb) 1616 { 1617 __sb_end_write(sb, SB_FREEZE_FS); 1618 } 1619 1620 /** 1621 * sb_start_write - get write access to a superblock 1622 * @sb: the super we write to 1623 * 1624 * When a process wants to write data or metadata to a file system (i.e. dirty 1625 * a page or an inode), it should embed the operation in a sb_start_write() - 1626 * sb_end_write() pair to get exclusion against file system freezing. This 1627 * function increments number of writers preventing freezing. If the file 1628 * system is already frozen, the function waits until the file system is 1629 * thawed. 1630 * 1631 * Since freeze protection behaves as a lock, users have to preserve 1632 * ordering of freeze protection and other filesystem locks. Generally, 1633 * freeze protection should be the outermost lock. In particular, we have: 1634 * 1635 * sb_start_write 1636 * -> i_mutex (write path, truncate, directory ops, ...) 1637 * -> s_umount (freeze_super, thaw_super) 1638 */ 1639 static inline void sb_start_write(struct super_block *sb) 1640 { 1641 __sb_start_write(sb, SB_FREEZE_WRITE); 1642 } 1643 1644 static inline bool sb_start_write_trylock(struct super_block *sb) 1645 { 1646 return __sb_start_write_trylock(sb, SB_FREEZE_WRITE); 1647 } 1648 1649 /** 1650 * sb_start_pagefault - get write access to a superblock from a page fault 1651 * @sb: the super we write to 1652 * 1653 * When a process starts handling write page fault, it should embed the 1654 * operation into sb_start_pagefault() - sb_end_pagefault() pair to get 1655 * exclusion against file system freezing. This is needed since the page fault 1656 * is going to dirty a page. This function increments number of running page 1657 * faults preventing freezing. If the file system is already frozen, the 1658 * function waits until the file system is thawed. 1659 * 1660 * Since page fault freeze protection behaves as a lock, users have to preserve 1661 * ordering of freeze protection and other filesystem locks. It is advised to 1662 * put sb_start_pagefault() close to mmap_lock in lock ordering. Page fault 1663 * handling code implies lock dependency: 1664 * 1665 * mmap_lock 1666 * -> sb_start_pagefault 1667 */ 1668 static inline void sb_start_pagefault(struct super_block *sb) 1669 { 1670 __sb_start_write(sb, SB_FREEZE_PAGEFAULT); 1671 } 1672 1673 /** 1674 * sb_start_intwrite - get write access to a superblock for internal fs purposes 1675 * @sb: the super we write to 1676 * 1677 * This is the third level of protection against filesystem freezing. It is 1678 * free for use by a filesystem. The only requirement is that it must rank 1679 * below sb_start_pagefault. 1680 * 1681 * For example filesystem can call sb_start_intwrite() when starting a 1682 * transaction which somewhat eases handling of freezing for internal sources 1683 * of filesystem changes (internal fs threads, discarding preallocation on file 1684 * close, etc.). 1685 */ 1686 static inline void sb_start_intwrite(struct super_block *sb) 1687 { 1688 __sb_start_write(sb, SB_FREEZE_FS); 1689 } 1690 1691 static inline bool sb_start_intwrite_trylock(struct super_block *sb) 1692 { 1693 return __sb_start_write_trylock(sb, SB_FREEZE_FS); 1694 } 1695 1696 bool inode_owner_or_capable(struct mnt_idmap *idmap, 1697 const struct inode *inode); 1698 1699 /* 1700 * VFS helper functions.. 1701 */ 1702 int vfs_create(struct mnt_idmap *, struct inode *, 1703 struct dentry *, umode_t, bool); 1704 int vfs_mkdir(struct mnt_idmap *, struct inode *, 1705 struct dentry *, umode_t); 1706 int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *, 1707 umode_t, dev_t); 1708 int vfs_symlink(struct mnt_idmap *, struct inode *, 1709 struct dentry *, const char *); 1710 int vfs_link(struct dentry *, struct mnt_idmap *, struct inode *, 1711 struct dentry *, struct inode **); 1712 int vfs_rmdir(struct mnt_idmap *, struct inode *, struct dentry *); 1713 int vfs_unlink(struct mnt_idmap *, struct inode *, struct dentry *, 1714 struct inode **); 1715 1716 /** 1717 * struct renamedata - contains all information required for renaming 1718 * @old_mnt_idmap: idmap of the old mount the inode was found from 1719 * @old_dir: parent of source 1720 * @old_dentry: source 1721 * @new_mnt_idmap: idmap of the new mount the inode was found from 1722 * @new_dir: parent of destination 1723 * @new_dentry: destination 1724 * @delegated_inode: returns an inode needing a delegation break 1725 * @flags: rename flags 1726 */ 1727 struct renamedata { 1728 struct mnt_idmap *old_mnt_idmap; 1729 struct inode *old_dir; 1730 struct dentry *old_dentry; 1731 struct mnt_idmap *new_mnt_idmap; 1732 struct inode *new_dir; 1733 struct dentry *new_dentry; 1734 struct inode **delegated_inode; 1735 unsigned int flags; 1736 } __randomize_layout; 1737 1738 int vfs_rename(struct renamedata *); 1739 1740 static inline int vfs_whiteout(struct mnt_idmap *idmap, 1741 struct inode *dir, struct dentry *dentry) 1742 { 1743 return vfs_mknod(idmap, dir, dentry, S_IFCHR | WHITEOUT_MODE, 1744 WHITEOUT_DEV); 1745 } 1746 1747 struct file *kernel_tmpfile_open(struct mnt_idmap *idmap, 1748 const struct path *parentpath, 1749 umode_t mode, int open_flag, 1750 const struct cred *cred); 1751 struct file *kernel_file_open(const struct path *path, int flags, 1752 struct inode *inode, const struct cred *cred); 1753 1754 int vfs_mkobj(struct dentry *, umode_t, 1755 int (*f)(struct dentry *, umode_t, void *), 1756 void *); 1757 1758 int vfs_fchown(struct file *file, uid_t user, gid_t group); 1759 int vfs_fchmod(struct file *file, umode_t mode); 1760 int vfs_utimes(const struct path *path, struct timespec64 *times); 1761 1762 extern long vfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 1763 1764 #ifdef CONFIG_COMPAT 1765 extern long compat_ptr_ioctl(struct file *file, unsigned int cmd, 1766 unsigned long arg); 1767 #else 1768 #define compat_ptr_ioctl NULL 1769 #endif 1770 1771 /* 1772 * VFS file helper functions. 1773 */ 1774 void inode_init_owner(struct mnt_idmap *idmap, struct inode *inode, 1775 const struct inode *dir, umode_t mode); 1776 extern bool may_open_dev(const struct path *path); 1777 umode_t mode_strip_sgid(struct mnt_idmap *idmap, 1778 const struct inode *dir, umode_t mode); 1779 1780 /* 1781 * This is the "filldir" function type, used by readdir() to let 1782 * the kernel specify what kind of dirent layout it wants to have. 1783 * This allows the kernel to read directories into kernel space or 1784 * to have different dirent layouts depending on the binary type. 1785 * Return 'true' to keep going and 'false' if there are no more entries. 1786 */ 1787 struct dir_context; 1788 typedef bool (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64, 1789 unsigned); 1790 1791 struct dir_context { 1792 filldir_t actor; 1793 loff_t pos; 1794 }; 1795 1796 /* 1797 * These flags let !MMU mmap() govern direct device mapping vs immediate 1798 * copying more easily for MAP_PRIVATE, especially for ROM filesystems. 1799 * 1800 * NOMMU_MAP_COPY: Copy can be mapped (MAP_PRIVATE) 1801 * NOMMU_MAP_DIRECT: Can be mapped directly (MAP_SHARED) 1802 * NOMMU_MAP_READ: Can be mapped for reading 1803 * NOMMU_MAP_WRITE: Can be mapped for writing 1804 * NOMMU_MAP_EXEC: Can be mapped for execution 1805 */ 1806 #define NOMMU_MAP_COPY 0x00000001 1807 #define NOMMU_MAP_DIRECT 0x00000008 1808 #define NOMMU_MAP_READ VM_MAYREAD 1809 #define NOMMU_MAP_WRITE VM_MAYWRITE 1810 #define NOMMU_MAP_EXEC VM_MAYEXEC 1811 1812 #define NOMMU_VMFLAGS \ 1813 (NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC) 1814 1815 /* 1816 * These flags control the behavior of the remap_file_range function pointer. 1817 * If it is called with len == 0 that means "remap to end of source file". 1818 * See Documentation/filesystems/vfs.rst for more details about this call. 1819 * 1820 * REMAP_FILE_DEDUP: only remap if contents identical (i.e. deduplicate) 1821 * REMAP_FILE_CAN_SHORTEN: caller can handle a shortened request 1822 */ 1823 #define REMAP_FILE_DEDUP (1 << 0) 1824 #define REMAP_FILE_CAN_SHORTEN (1 << 1) 1825 1826 /* 1827 * These flags signal that the caller is ok with altering various aspects of 1828 * the behavior of the remap operation. The changes must be made by the 1829 * implementation; the vfs remap helper functions can take advantage of them. 1830 * Flags in this category exist to preserve the quirky behavior of the hoisted 1831 * btrfs clone/dedupe ioctls. 1832 */ 1833 #define REMAP_FILE_ADVISORY (REMAP_FILE_CAN_SHORTEN) 1834 1835 /* 1836 * These flags control the behavior of vfs_copy_file_range(). 1837 * They are not available to the user via syscall. 1838 * 1839 * COPY_FILE_SPLICE: call splice direct instead of fs clone/copy ops 1840 */ 1841 #define COPY_FILE_SPLICE (1 << 0) 1842 1843 struct iov_iter; 1844 struct io_uring_cmd; 1845 1846 struct file_operations { 1847 struct module *owner; 1848 loff_t (*llseek) (struct file *, loff_t, int); 1849 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *); 1850 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *); 1851 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *); 1852 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *); 1853 int (*iopoll)(struct kiocb *kiocb, struct io_comp_batch *, 1854 unsigned int flags); 1855 int (*iterate) (struct file *, struct dir_context *); 1856 int (*iterate_shared) (struct file *, struct dir_context *); 1857 __poll_t (*poll) (struct file *, struct poll_table_struct *); 1858 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long); 1859 long (*compat_ioctl) (struct file *, unsigned int, unsigned long); 1860 int (*mmap) (struct file *, struct vm_area_struct *); 1861 unsigned long mmap_supported_flags; 1862 int (*open) (struct inode *, struct file *); 1863 int (*flush) (struct file *, fl_owner_t id); 1864 int (*release) (struct inode *, struct file *); 1865 int (*fsync) (struct file *, loff_t, loff_t, int datasync); 1866 int (*fasync) (int, struct file *, int); 1867 int (*lock) (struct file *, int, struct file_lock *); 1868 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 1869 int (*check_flags)(int); 1870 int (*flock) (struct file *, int, struct file_lock *); 1871 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int); 1872 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int); 1873 void (*splice_eof)(struct file *file); 1874 int (*setlease)(struct file *, long, struct file_lock **, void **); 1875 long (*fallocate)(struct file *file, int mode, loff_t offset, 1876 loff_t len); 1877 void (*show_fdinfo)(struct seq_file *m, struct file *f); 1878 #ifndef CONFIG_MMU 1879 unsigned (*mmap_capabilities)(struct file *); 1880 #endif 1881 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *, 1882 loff_t, size_t, unsigned int); 1883 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in, 1884 struct file *file_out, loff_t pos_out, 1885 loff_t len, unsigned int remap_flags); 1886 int (*fadvise)(struct file *, loff_t, loff_t, int); 1887 int (*uring_cmd)(struct io_uring_cmd *ioucmd, unsigned int issue_flags); 1888 int (*uring_cmd_iopoll)(struct io_uring_cmd *, struct io_comp_batch *, 1889 unsigned int poll_flags); 1890 } __randomize_layout; 1891 1892 struct inode_operations { 1893 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int); 1894 const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *); 1895 int (*permission) (struct mnt_idmap *, struct inode *, int); 1896 struct posix_acl * (*get_inode_acl)(struct inode *, int, bool); 1897 1898 int (*readlink) (struct dentry *, char __user *,int); 1899 1900 int (*create) (struct mnt_idmap *, struct inode *,struct dentry *, 1901 umode_t, bool); 1902 int (*link) (struct dentry *,struct inode *,struct dentry *); 1903 int (*unlink) (struct inode *,struct dentry *); 1904 int (*symlink) (struct mnt_idmap *, struct inode *,struct dentry *, 1905 const char *); 1906 int (*mkdir) (struct mnt_idmap *, struct inode *,struct dentry *, 1907 umode_t); 1908 int (*rmdir) (struct inode *,struct dentry *); 1909 int (*mknod) (struct mnt_idmap *, struct inode *,struct dentry *, 1910 umode_t,dev_t); 1911 int (*rename) (struct mnt_idmap *, struct inode *, struct dentry *, 1912 struct inode *, struct dentry *, unsigned int); 1913 int (*setattr) (struct mnt_idmap *, struct dentry *, struct iattr *); 1914 int (*getattr) (struct mnt_idmap *, const struct path *, 1915 struct kstat *, u32, unsigned int); 1916 ssize_t (*listxattr) (struct dentry *, char *, size_t); 1917 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start, 1918 u64 len); 1919 int (*update_time)(struct inode *, int); 1920 int (*atomic_open)(struct inode *, struct dentry *, 1921 struct file *, unsigned open_flag, 1922 umode_t create_mode); 1923 int (*tmpfile) (struct mnt_idmap *, struct inode *, 1924 struct file *, umode_t); 1925 struct posix_acl *(*get_acl)(struct mnt_idmap *, struct dentry *, 1926 int); 1927 int (*set_acl)(struct mnt_idmap *, struct dentry *, 1928 struct posix_acl *, int); 1929 int (*fileattr_set)(struct mnt_idmap *idmap, 1930 struct dentry *dentry, struct fileattr *fa); 1931 int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa); 1932 } ____cacheline_aligned; 1933 1934 static inline ssize_t call_read_iter(struct file *file, struct kiocb *kio, 1935 struct iov_iter *iter) 1936 { 1937 return file->f_op->read_iter(kio, iter); 1938 } 1939 1940 static inline ssize_t call_write_iter(struct file *file, struct kiocb *kio, 1941 struct iov_iter *iter) 1942 { 1943 return file->f_op->write_iter(kio, iter); 1944 } 1945 1946 static inline int call_mmap(struct file *file, struct vm_area_struct *vma) 1947 { 1948 return file->f_op->mmap(file, vma); 1949 } 1950 1951 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *); 1952 extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *); 1953 extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *, 1954 loff_t, size_t, unsigned int); 1955 extern ssize_t generic_copy_file_range(struct file *file_in, loff_t pos_in, 1956 struct file *file_out, loff_t pos_out, 1957 size_t len, unsigned int flags); 1958 int __generic_remap_file_range_prep(struct file *file_in, loff_t pos_in, 1959 struct file *file_out, loff_t pos_out, 1960 loff_t *len, unsigned int remap_flags, 1961 const struct iomap_ops *dax_read_ops); 1962 int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in, 1963 struct file *file_out, loff_t pos_out, 1964 loff_t *count, unsigned int remap_flags); 1965 extern loff_t do_clone_file_range(struct file *file_in, loff_t pos_in, 1966 struct file *file_out, loff_t pos_out, 1967 loff_t len, unsigned int remap_flags); 1968 extern loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in, 1969 struct file *file_out, loff_t pos_out, 1970 loff_t len, unsigned int remap_flags); 1971 extern int vfs_dedupe_file_range(struct file *file, 1972 struct file_dedupe_range *same); 1973 extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos, 1974 struct file *dst_file, loff_t dst_pos, 1975 loff_t len, unsigned int remap_flags); 1976 1977 1978 struct super_operations { 1979 struct inode *(*alloc_inode)(struct super_block *sb); 1980 void (*destroy_inode)(struct inode *); 1981 void (*free_inode)(struct inode *); 1982 1983 void (*dirty_inode) (struct inode *, int flags); 1984 int (*write_inode) (struct inode *, struct writeback_control *wbc); 1985 int (*drop_inode) (struct inode *); 1986 void (*evict_inode) (struct inode *); 1987 void (*put_super) (struct super_block *); 1988 int (*sync_fs)(struct super_block *sb, int wait); 1989 int (*freeze_super) (struct super_block *); 1990 int (*freeze_fs) (struct super_block *); 1991 int (*thaw_super) (struct super_block *); 1992 int (*unfreeze_fs) (struct super_block *); 1993 int (*statfs) (struct dentry *, struct kstatfs *); 1994 int (*remount_fs) (struct super_block *, int *, char *); 1995 void (*umount_begin) (struct super_block *); 1996 1997 int (*show_options)(struct seq_file *, struct dentry *); 1998 int (*show_devname)(struct seq_file *, struct dentry *); 1999 int (*show_path)(struct seq_file *, struct dentry *); 2000 int (*show_stats)(struct seq_file *, struct dentry *); 2001 #ifdef CONFIG_QUOTA 2002 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t); 2003 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t); 2004 struct dquot **(*get_dquots)(struct inode *); 2005 #endif 2006 long (*nr_cached_objects)(struct super_block *, 2007 struct shrink_control *); 2008 long (*free_cached_objects)(struct super_block *, 2009 struct shrink_control *); 2010 void (*shutdown)(struct super_block *sb); 2011 }; 2012 2013 /* 2014 * Inode flags - they have no relation to superblock flags now 2015 */ 2016 #define S_SYNC (1 << 0) /* Writes are synced at once */ 2017 #define S_NOATIME (1 << 1) /* Do not update access times */ 2018 #define S_APPEND (1 << 2) /* Append-only file */ 2019 #define S_IMMUTABLE (1 << 3) /* Immutable file */ 2020 #define S_DEAD (1 << 4) /* removed, but still open directory */ 2021 #define S_NOQUOTA (1 << 5) /* Inode is not counted to quota */ 2022 #define S_DIRSYNC (1 << 6) /* Directory modifications are synchronous */ 2023 #define S_NOCMTIME (1 << 7) /* Do not update file c/mtime */ 2024 #define S_SWAPFILE (1 << 8) /* Do not truncate: swapon got its bmaps */ 2025 #define S_PRIVATE (1 << 9) /* Inode is fs-internal */ 2026 #define S_IMA (1 << 10) /* Inode has an associated IMA struct */ 2027 #define S_AUTOMOUNT (1 << 11) /* Automount/referral quasi-directory */ 2028 #define S_NOSEC (1 << 12) /* no suid or xattr security attributes */ 2029 #ifdef CONFIG_FS_DAX 2030 #define S_DAX (1 << 13) /* Direct Access, avoiding the page cache */ 2031 #else 2032 #define S_DAX 0 /* Make all the DAX code disappear */ 2033 #endif 2034 #define S_ENCRYPTED (1 << 14) /* Encrypted file (using fs/crypto/) */ 2035 #define S_CASEFOLD (1 << 15) /* Casefolded file */ 2036 #define S_VERITY (1 << 16) /* Verity file (using fs/verity/) */ 2037 #define S_KERNEL_FILE (1 << 17) /* File is in use by the kernel (eg. fs/cachefiles) */ 2038 2039 /* 2040 * Note that nosuid etc flags are inode-specific: setting some file-system 2041 * flags just means all the inodes inherit those flags by default. It might be 2042 * possible to override it selectively if you really wanted to with some 2043 * ioctl() that is not currently implemented. 2044 * 2045 * Exception: SB_RDONLY is always applied to the entire file system. 2046 * 2047 * Unfortunately, it is possible to change a filesystems flags with it mounted 2048 * with files in use. This means that all of the inodes will not have their 2049 * i_flags updated. Hence, i_flags no longer inherit the superblock mount 2050 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org 2051 */ 2052 #define __IS_FLG(inode, flg) ((inode)->i_sb->s_flags & (flg)) 2053 2054 static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; } 2055 #define IS_RDONLY(inode) sb_rdonly((inode)->i_sb) 2056 #define IS_SYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS) || \ 2057 ((inode)->i_flags & S_SYNC)) 2058 #define IS_DIRSYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS|SB_DIRSYNC) || \ 2059 ((inode)->i_flags & (S_SYNC|S_DIRSYNC))) 2060 #define IS_MANDLOCK(inode) __IS_FLG(inode, SB_MANDLOCK) 2061 #define IS_NOATIME(inode) __IS_FLG(inode, SB_RDONLY|SB_NOATIME) 2062 #define IS_I_VERSION(inode) __IS_FLG(inode, SB_I_VERSION) 2063 2064 #define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA) 2065 #define IS_APPEND(inode) ((inode)->i_flags & S_APPEND) 2066 #define IS_IMMUTABLE(inode) ((inode)->i_flags & S_IMMUTABLE) 2067 #define IS_POSIXACL(inode) __IS_FLG(inode, SB_POSIXACL) 2068 2069 #define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD) 2070 #define IS_NOCMTIME(inode) ((inode)->i_flags & S_NOCMTIME) 2071 #define IS_SWAPFILE(inode) ((inode)->i_flags & S_SWAPFILE) 2072 #define IS_PRIVATE(inode) ((inode)->i_flags & S_PRIVATE) 2073 #define IS_IMA(inode) ((inode)->i_flags & S_IMA) 2074 #define IS_AUTOMOUNT(inode) ((inode)->i_flags & S_AUTOMOUNT) 2075 #define IS_NOSEC(inode) ((inode)->i_flags & S_NOSEC) 2076 #define IS_DAX(inode) ((inode)->i_flags & S_DAX) 2077 #define IS_ENCRYPTED(inode) ((inode)->i_flags & S_ENCRYPTED) 2078 #define IS_CASEFOLDED(inode) ((inode)->i_flags & S_CASEFOLD) 2079 #define IS_VERITY(inode) ((inode)->i_flags & S_VERITY) 2080 2081 #define IS_WHITEOUT(inode) (S_ISCHR(inode->i_mode) && \ 2082 (inode)->i_rdev == WHITEOUT_DEV) 2083 2084 static inline bool HAS_UNMAPPED_ID(struct mnt_idmap *idmap, 2085 struct inode *inode) 2086 { 2087 return !vfsuid_valid(i_uid_into_vfsuid(idmap, inode)) || 2088 !vfsgid_valid(i_gid_into_vfsgid(idmap, inode)); 2089 } 2090 2091 static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp) 2092 { 2093 *kiocb = (struct kiocb) { 2094 .ki_filp = filp, 2095 .ki_flags = filp->f_iocb_flags, 2096 .ki_ioprio = get_current_ioprio(), 2097 }; 2098 } 2099 2100 static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src, 2101 struct file *filp) 2102 { 2103 *kiocb = (struct kiocb) { 2104 .ki_filp = filp, 2105 .ki_flags = kiocb_src->ki_flags, 2106 .ki_ioprio = kiocb_src->ki_ioprio, 2107 .ki_pos = kiocb_src->ki_pos, 2108 }; 2109 } 2110 2111 /* 2112 * Inode state bits. Protected by inode->i_lock 2113 * 2114 * Four bits determine the dirty state of the inode: I_DIRTY_SYNC, 2115 * I_DIRTY_DATASYNC, I_DIRTY_PAGES, and I_DIRTY_TIME. 2116 * 2117 * Four bits define the lifetime of an inode. Initially, inodes are I_NEW, 2118 * until that flag is cleared. I_WILL_FREE, I_FREEING and I_CLEAR are set at 2119 * various stages of removing an inode. 2120 * 2121 * Two bits are used for locking and completion notification, I_NEW and I_SYNC. 2122 * 2123 * I_DIRTY_SYNC Inode is dirty, but doesn't have to be written on 2124 * fdatasync() (unless I_DIRTY_DATASYNC is also set). 2125 * Timestamp updates are the usual cause. 2126 * I_DIRTY_DATASYNC Data-related inode changes pending. We keep track of 2127 * these changes separately from I_DIRTY_SYNC so that we 2128 * don't have to write inode on fdatasync() when only 2129 * e.g. the timestamps have changed. 2130 * I_DIRTY_PAGES Inode has dirty pages. Inode itself may be clean. 2131 * I_DIRTY_TIME The inode itself has dirty timestamps, and the 2132 * lazytime mount option is enabled. We keep track of this 2133 * separately from I_DIRTY_SYNC in order to implement 2134 * lazytime. This gets cleared if I_DIRTY_INODE 2135 * (I_DIRTY_SYNC and/or I_DIRTY_DATASYNC) gets set. But 2136 * I_DIRTY_TIME can still be set if I_DIRTY_SYNC is already 2137 * in place because writeback might already be in progress 2138 * and we don't want to lose the time update 2139 * I_NEW Serves as both a mutex and completion notification. 2140 * New inodes set I_NEW. If two processes both create 2141 * the same inode, one of them will release its inode and 2142 * wait for I_NEW to be released before returning. 2143 * Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can 2144 * also cause waiting on I_NEW, without I_NEW actually 2145 * being set. find_inode() uses this to prevent returning 2146 * nearly-dead inodes. 2147 * I_WILL_FREE Must be set when calling write_inode_now() if i_count 2148 * is zero. I_FREEING must be set when I_WILL_FREE is 2149 * cleared. 2150 * I_FREEING Set when inode is about to be freed but still has dirty 2151 * pages or buffers attached or the inode itself is still 2152 * dirty. 2153 * I_CLEAR Added by clear_inode(). In this state the inode is 2154 * clean and can be destroyed. Inode keeps I_FREEING. 2155 * 2156 * Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are 2157 * prohibited for many purposes. iget() must wait for 2158 * the inode to be completely released, then create it 2159 * anew. Other functions will just ignore such inodes, 2160 * if appropriate. I_NEW is used for waiting. 2161 * 2162 * I_SYNC Writeback of inode is running. The bit is set during 2163 * data writeback, and cleared with a wakeup on the bit 2164 * address once it is done. The bit is also used to pin 2165 * the inode in memory for flusher thread. 2166 * 2167 * I_REFERENCED Marks the inode as recently references on the LRU list. 2168 * 2169 * I_DIO_WAKEUP Never set. Only used as a key for wait_on_bit(). 2170 * 2171 * I_WB_SWITCH Cgroup bdi_writeback switching in progress. Used to 2172 * synchronize competing switching instances and to tell 2173 * wb stat updates to grab the i_pages lock. See 2174 * inode_switch_wbs_work_fn() for details. 2175 * 2176 * I_OVL_INUSE Used by overlayfs to get exclusive ownership on upper 2177 * and work dirs among overlayfs mounts. 2178 * 2179 * I_CREATING New object's inode in the middle of setting up. 2180 * 2181 * I_DONTCACHE Evict inode as soon as it is not used anymore. 2182 * 2183 * I_SYNC_QUEUED Inode is queued in b_io or b_more_io writeback lists. 2184 * Used to detect that mark_inode_dirty() should not move 2185 * inode between dirty lists. 2186 * 2187 * I_PINNING_FSCACHE_WB Inode is pinning an fscache object for writeback. 2188 * 2189 * Q: What is the difference between I_WILL_FREE and I_FREEING? 2190 */ 2191 #define I_DIRTY_SYNC (1 << 0) 2192 #define I_DIRTY_DATASYNC (1 << 1) 2193 #define I_DIRTY_PAGES (1 << 2) 2194 #define __I_NEW 3 2195 #define I_NEW (1 << __I_NEW) 2196 #define I_WILL_FREE (1 << 4) 2197 #define I_FREEING (1 << 5) 2198 #define I_CLEAR (1 << 6) 2199 #define __I_SYNC 7 2200 #define I_SYNC (1 << __I_SYNC) 2201 #define I_REFERENCED (1 << 8) 2202 #define __I_DIO_WAKEUP 9 2203 #define I_DIO_WAKEUP (1 << __I_DIO_WAKEUP) 2204 #define I_LINKABLE (1 << 10) 2205 #define I_DIRTY_TIME (1 << 11) 2206 #define I_WB_SWITCH (1 << 13) 2207 #define I_OVL_INUSE (1 << 14) 2208 #define I_CREATING (1 << 15) 2209 #define I_DONTCACHE (1 << 16) 2210 #define I_SYNC_QUEUED (1 << 17) 2211 #define I_PINNING_FSCACHE_WB (1 << 18) 2212 2213 #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC) 2214 #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES) 2215 #define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME) 2216 2217 extern void __mark_inode_dirty(struct inode *, int); 2218 static inline void mark_inode_dirty(struct inode *inode) 2219 { 2220 __mark_inode_dirty(inode, I_DIRTY); 2221 } 2222 2223 static inline void mark_inode_dirty_sync(struct inode *inode) 2224 { 2225 __mark_inode_dirty(inode, I_DIRTY_SYNC); 2226 } 2227 2228 /* 2229 * Returns true if the given inode itself only has dirty timestamps (its pages 2230 * may still be dirty) and isn't currently being allocated or freed. 2231 * Filesystems should call this if when writing an inode when lazytime is 2232 * enabled, they want to opportunistically write the timestamps of other inodes 2233 * located very nearby on-disk, e.g. in the same inode block. This returns true 2234 * if the given inode is in need of such an opportunistic update. Requires 2235 * i_lock, or at least later re-checking under i_lock. 2236 */ 2237 static inline bool inode_is_dirtytime_only(struct inode *inode) 2238 { 2239 return (inode->i_state & (I_DIRTY_TIME | I_NEW | 2240 I_FREEING | I_WILL_FREE)) == I_DIRTY_TIME; 2241 } 2242 2243 extern void inc_nlink(struct inode *inode); 2244 extern void drop_nlink(struct inode *inode); 2245 extern void clear_nlink(struct inode *inode); 2246 extern void set_nlink(struct inode *inode, unsigned int nlink); 2247 2248 static inline void inode_inc_link_count(struct inode *inode) 2249 { 2250 inc_nlink(inode); 2251 mark_inode_dirty(inode); 2252 } 2253 2254 static inline void inode_dec_link_count(struct inode *inode) 2255 { 2256 drop_nlink(inode); 2257 mark_inode_dirty(inode); 2258 } 2259 2260 enum file_time_flags { 2261 S_ATIME = 1, 2262 S_MTIME = 2, 2263 S_CTIME = 4, 2264 S_VERSION = 8, 2265 }; 2266 2267 extern bool atime_needs_update(const struct path *, struct inode *); 2268 extern void touch_atime(const struct path *); 2269 int inode_update_time(struct inode *inode, int flags); 2270 2271 static inline void file_accessed(struct file *file) 2272 { 2273 if (!(file->f_flags & O_NOATIME)) 2274 touch_atime(&file->f_path); 2275 } 2276 2277 extern int file_modified(struct file *file); 2278 int kiocb_modified(struct kiocb *iocb); 2279 2280 int sync_inode_metadata(struct inode *inode, int wait); 2281 2282 struct file_system_type { 2283 const char *name; 2284 int fs_flags; 2285 #define FS_REQUIRES_DEV 1 2286 #define FS_BINARY_MOUNTDATA 2 2287 #define FS_HAS_SUBTYPE 4 2288 #define FS_USERNS_MOUNT 8 /* Can be mounted by userns root */ 2289 #define FS_DISALLOW_NOTIFY_PERM 16 /* Disable fanotify permission events */ 2290 #define FS_ALLOW_IDMAP 32 /* FS has been updated to handle vfs idmappings. */ 2291 #define FS_MGTIME 64 /* FS uses multigrain timestamps */ 2292 #define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() during rename() internally. */ 2293 int (*init_fs_context)(struct fs_context *); 2294 const struct fs_parameter_spec *parameters; 2295 struct dentry *(*mount) (struct file_system_type *, int, 2296 const char *, void *); 2297 void (*kill_sb) (struct super_block *); 2298 struct module *owner; 2299 struct file_system_type * next; 2300 struct hlist_head fs_supers; 2301 2302 struct lock_class_key s_lock_key; 2303 struct lock_class_key s_umount_key; 2304 struct lock_class_key s_vfs_rename_key; 2305 struct lock_class_key s_writers_key[SB_FREEZE_LEVELS]; 2306 2307 struct lock_class_key i_lock_key; 2308 struct lock_class_key i_mutex_key; 2309 struct lock_class_key invalidate_lock_key; 2310 struct lock_class_key i_mutex_dir_key; 2311 }; 2312 2313 #define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME) 2314 2315 /** 2316 * is_mgtime: is this inode using multigrain timestamps 2317 * @inode: inode to test for multigrain timestamps 2318 * 2319 * Return true if the inode uses multigrain timestamps, false otherwise. 2320 */ 2321 static inline bool is_mgtime(const struct inode *inode) 2322 { 2323 return inode->i_sb->s_type->fs_flags & FS_MGTIME; 2324 } 2325 2326 extern struct dentry *mount_bdev(struct file_system_type *fs_type, 2327 int flags, const char *dev_name, void *data, 2328 int (*fill_super)(struct super_block *, void *, int)); 2329 extern struct dentry *mount_single(struct file_system_type *fs_type, 2330 int flags, void *data, 2331 int (*fill_super)(struct super_block *, void *, int)); 2332 extern struct dentry *mount_nodev(struct file_system_type *fs_type, 2333 int flags, void *data, 2334 int (*fill_super)(struct super_block *, void *, int)); 2335 extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path); 2336 void retire_super(struct super_block *sb); 2337 void generic_shutdown_super(struct super_block *sb); 2338 void kill_block_super(struct super_block *sb); 2339 void kill_anon_super(struct super_block *sb); 2340 void kill_litter_super(struct super_block *sb); 2341 void deactivate_super(struct super_block *sb); 2342 void deactivate_locked_super(struct super_block *sb); 2343 int set_anon_super(struct super_block *s, void *data); 2344 int set_anon_super_fc(struct super_block *s, struct fs_context *fc); 2345 int get_anon_bdev(dev_t *); 2346 void free_anon_bdev(dev_t); 2347 struct super_block *sget_fc(struct fs_context *fc, 2348 int (*test)(struct super_block *, struct fs_context *), 2349 int (*set)(struct super_block *, struct fs_context *)); 2350 struct super_block *sget(struct file_system_type *type, 2351 int (*test)(struct super_block *,void *), 2352 int (*set)(struct super_block *,void *), 2353 int flags, void *data); 2354 2355 /* Alas, no aliases. Too much hassle with bringing module.h everywhere */ 2356 #define fops_get(fops) \ 2357 (((fops) && try_module_get((fops)->owner) ? (fops) : NULL)) 2358 #define fops_put(fops) \ 2359 do { if (fops) module_put((fops)->owner); } while(0) 2360 /* 2361 * This one is to be used *ONLY* from ->open() instances. 2362 * fops must be non-NULL, pinned down *and* module dependencies 2363 * should be sufficient to pin the caller down as well. 2364 */ 2365 #define replace_fops(f, fops) \ 2366 do { \ 2367 struct file *__file = (f); \ 2368 fops_put(__file->f_op); \ 2369 BUG_ON(!(__file->f_op = (fops))); \ 2370 } while(0) 2371 2372 extern int register_filesystem(struct file_system_type *); 2373 extern int unregister_filesystem(struct file_system_type *); 2374 extern int vfs_statfs(const struct path *, struct kstatfs *); 2375 extern int user_statfs(const char __user *, struct kstatfs *); 2376 extern int fd_statfs(int, struct kstatfs *); 2377 extern int freeze_super(struct super_block *super); 2378 extern int thaw_super(struct super_block *super); 2379 extern __printf(2, 3) 2380 int super_setup_bdi_name(struct super_block *sb, char *fmt, ...); 2381 extern int super_setup_bdi(struct super_block *sb); 2382 2383 extern int current_umask(void); 2384 2385 extern void ihold(struct inode * inode); 2386 extern void iput(struct inode *); 2387 int inode_update_timestamps(struct inode *inode, int flags); 2388 int generic_update_time(struct inode *, int); 2389 2390 /* /sys/fs */ 2391 extern struct kobject *fs_kobj; 2392 2393 #define MAX_RW_COUNT (INT_MAX & PAGE_MASK) 2394 2395 /* fs/open.c */ 2396 struct audit_names; 2397 struct filename { 2398 const char *name; /* pointer to actual string */ 2399 const __user char *uptr; /* original userland pointer */ 2400 int refcnt; 2401 struct audit_names *aname; 2402 const char iname[]; 2403 }; 2404 static_assert(offsetof(struct filename, iname) % sizeof(long) == 0); 2405 2406 static inline struct mnt_idmap *file_mnt_idmap(struct file *file) 2407 { 2408 return mnt_idmap(file->f_path.mnt); 2409 } 2410 2411 /** 2412 * is_idmapped_mnt - check whether a mount is mapped 2413 * @mnt: the mount to check 2414 * 2415 * If @mnt has an non @nop_mnt_idmap attached to it then @mnt is mapped. 2416 * 2417 * Return: true if mount is mapped, false if not. 2418 */ 2419 static inline bool is_idmapped_mnt(const struct vfsmount *mnt) 2420 { 2421 return mnt_idmap(mnt) != &nop_mnt_idmap; 2422 } 2423 2424 extern long vfs_truncate(const struct path *, loff_t); 2425 int do_truncate(struct mnt_idmap *, struct dentry *, loff_t start, 2426 unsigned int time_attrs, struct file *filp); 2427 extern int vfs_fallocate(struct file *file, int mode, loff_t offset, 2428 loff_t len); 2429 extern long do_sys_open(int dfd, const char __user *filename, int flags, 2430 umode_t mode); 2431 extern struct file *file_open_name(struct filename *, int, umode_t); 2432 extern struct file *filp_open(const char *, int, umode_t); 2433 extern struct file *file_open_root(const struct path *, 2434 const char *, int, umode_t); 2435 static inline struct file *file_open_root_mnt(struct vfsmount *mnt, 2436 const char *name, int flags, umode_t mode) 2437 { 2438 return file_open_root(&(struct path){.mnt = mnt, .dentry = mnt->mnt_root}, 2439 name, flags, mode); 2440 } 2441 struct file *dentry_open(const struct path *path, int flags, 2442 const struct cred *creds); 2443 struct file *dentry_create(const struct path *path, int flags, umode_t mode, 2444 const struct cred *cred); 2445 struct file *backing_file_open(const struct path *path, int flags, 2446 const struct path *real_path, 2447 const struct cred *cred); 2448 struct path *backing_file_real_path(struct file *f); 2449 2450 /* 2451 * file_real_path - get the path corresponding to f_inode 2452 * 2453 * When opening a backing file for a stackable filesystem (e.g., 2454 * overlayfs) f_path may be on the stackable filesystem and f_inode on 2455 * the underlying filesystem. When the path associated with f_inode is 2456 * needed, this helper should be used instead of accessing f_path 2457 * directly. 2458 */ 2459 static inline const struct path *file_real_path(struct file *f) 2460 { 2461 if (unlikely(f->f_mode & FMODE_BACKING)) 2462 return backing_file_real_path(f); 2463 return &f->f_path; 2464 } 2465 2466 static inline struct file *file_clone_open(struct file *file) 2467 { 2468 return dentry_open(&file->f_path, file->f_flags, file->f_cred); 2469 } 2470 extern int filp_close(struct file *, fl_owner_t id); 2471 2472 extern struct filename *getname_flags(const char __user *, int, int *); 2473 extern struct filename *getname_uflags(const char __user *, int); 2474 extern struct filename *getname(const char __user *); 2475 extern struct filename *getname_kernel(const char *); 2476 extern void putname(struct filename *name); 2477 2478 extern int finish_open(struct file *file, struct dentry *dentry, 2479 int (*open)(struct inode *, struct file *)); 2480 extern int finish_no_open(struct file *file, struct dentry *dentry); 2481 2482 /* Helper for the simple case when original dentry is used */ 2483 static inline int finish_open_simple(struct file *file, int error) 2484 { 2485 if (error) 2486 return error; 2487 2488 return finish_open(file, file->f_path.dentry, NULL); 2489 } 2490 2491 /* fs/dcache.c */ 2492 extern void __init vfs_caches_init_early(void); 2493 extern void __init vfs_caches_init(void); 2494 2495 extern struct kmem_cache *names_cachep; 2496 2497 #define __getname() kmem_cache_alloc(names_cachep, GFP_KERNEL) 2498 #define __putname(name) kmem_cache_free(names_cachep, (void *)(name)) 2499 2500 extern struct super_block *blockdev_superblock; 2501 static inline bool sb_is_blkdev_sb(struct super_block *sb) 2502 { 2503 return IS_ENABLED(CONFIG_BLOCK) && sb == blockdev_superblock; 2504 } 2505 2506 void emergency_thaw_all(void); 2507 extern int sync_filesystem(struct super_block *); 2508 extern const struct file_operations def_blk_fops; 2509 extern const struct file_operations def_chr_fops; 2510 2511 /* fs/char_dev.c */ 2512 #define CHRDEV_MAJOR_MAX 512 2513 /* Marks the bottom of the first segment of free char majors */ 2514 #define CHRDEV_MAJOR_DYN_END 234 2515 /* Marks the top and bottom of the second segment of free char majors */ 2516 #define CHRDEV_MAJOR_DYN_EXT_START 511 2517 #define CHRDEV_MAJOR_DYN_EXT_END 384 2518 2519 extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *); 2520 extern int register_chrdev_region(dev_t, unsigned, const char *); 2521 extern int __register_chrdev(unsigned int major, unsigned int baseminor, 2522 unsigned int count, const char *name, 2523 const struct file_operations *fops); 2524 extern void __unregister_chrdev(unsigned int major, unsigned int baseminor, 2525 unsigned int count, const char *name); 2526 extern void unregister_chrdev_region(dev_t, unsigned); 2527 extern void chrdev_show(struct seq_file *,off_t); 2528 2529 static inline int register_chrdev(unsigned int major, const char *name, 2530 const struct file_operations *fops) 2531 { 2532 return __register_chrdev(major, 0, 256, name, fops); 2533 } 2534 2535 static inline void unregister_chrdev(unsigned int major, const char *name) 2536 { 2537 __unregister_chrdev(major, 0, 256, name); 2538 } 2539 2540 extern void init_special_inode(struct inode *, umode_t, dev_t); 2541 2542 /* Invalid inode operations -- fs/bad_inode.c */ 2543 extern void make_bad_inode(struct inode *); 2544 extern bool is_bad_inode(struct inode *); 2545 2546 extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart, 2547 loff_t lend); 2548 extern int __must_check file_check_and_advance_wb_err(struct file *file); 2549 extern int __must_check file_write_and_wait_range(struct file *file, 2550 loff_t start, loff_t end); 2551 2552 static inline int file_write_and_wait(struct file *file) 2553 { 2554 return file_write_and_wait_range(file, 0, LLONG_MAX); 2555 } 2556 2557 extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end, 2558 int datasync); 2559 extern int vfs_fsync(struct file *file, int datasync); 2560 2561 extern int sync_file_range(struct file *file, loff_t offset, loff_t nbytes, 2562 unsigned int flags); 2563 2564 static inline bool iocb_is_dsync(const struct kiocb *iocb) 2565 { 2566 return (iocb->ki_flags & IOCB_DSYNC) || 2567 IS_SYNC(iocb->ki_filp->f_mapping->host); 2568 } 2569 2570 /* 2571 * Sync the bytes written if this was a synchronous write. Expect ki_pos 2572 * to already be updated for the write, and will return either the amount 2573 * of bytes passed in, or an error if syncing the file failed. 2574 */ 2575 static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count) 2576 { 2577 if (iocb_is_dsync(iocb)) { 2578 int ret = vfs_fsync_range(iocb->ki_filp, 2579 iocb->ki_pos - count, iocb->ki_pos - 1, 2580 (iocb->ki_flags & IOCB_SYNC) ? 0 : 1); 2581 if (ret) 2582 return ret; 2583 } 2584 2585 return count; 2586 } 2587 2588 extern void emergency_sync(void); 2589 extern void emergency_remount(void); 2590 2591 #ifdef CONFIG_BLOCK 2592 extern int bmap(struct inode *inode, sector_t *block); 2593 #else 2594 static inline int bmap(struct inode *inode, sector_t *block) 2595 { 2596 return -EINVAL; 2597 } 2598 #endif 2599 2600 int notify_change(struct mnt_idmap *, struct dentry *, 2601 struct iattr *, struct inode **); 2602 int inode_permission(struct mnt_idmap *, struct inode *, int); 2603 int generic_permission(struct mnt_idmap *, struct inode *, int); 2604 static inline int file_permission(struct file *file, int mask) 2605 { 2606 return inode_permission(file_mnt_idmap(file), 2607 file_inode(file), mask); 2608 } 2609 static inline int path_permission(const struct path *path, int mask) 2610 { 2611 return inode_permission(mnt_idmap(path->mnt), 2612 d_inode(path->dentry), mask); 2613 } 2614 int __check_sticky(struct mnt_idmap *idmap, struct inode *dir, 2615 struct inode *inode); 2616 2617 static inline bool execute_ok(struct inode *inode) 2618 { 2619 return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode); 2620 } 2621 2622 static inline bool inode_wrong_type(const struct inode *inode, umode_t mode) 2623 { 2624 return (inode->i_mode ^ mode) & S_IFMT; 2625 } 2626 2627 static inline void file_start_write(struct file *file) 2628 { 2629 if (!S_ISREG(file_inode(file)->i_mode)) 2630 return; 2631 sb_start_write(file_inode(file)->i_sb); 2632 } 2633 2634 static inline bool file_start_write_trylock(struct file *file) 2635 { 2636 if (!S_ISREG(file_inode(file)->i_mode)) 2637 return true; 2638 return sb_start_write_trylock(file_inode(file)->i_sb); 2639 } 2640 2641 static inline void file_end_write(struct file *file) 2642 { 2643 if (!S_ISREG(file_inode(file)->i_mode)) 2644 return; 2645 __sb_end_write(file_inode(file)->i_sb, SB_FREEZE_WRITE); 2646 } 2647 2648 /* 2649 * This is used for regular files where some users -- especially the 2650 * currently executed binary in a process, previously handled via 2651 * VM_DENYWRITE -- cannot handle concurrent write (and maybe mmap 2652 * read-write shared) accesses. 2653 * 2654 * get_write_access() gets write permission for a file. 2655 * put_write_access() releases this write permission. 2656 * deny_write_access() denies write access to a file. 2657 * allow_write_access() re-enables write access to a file. 2658 * 2659 * The i_writecount field of an inode can have the following values: 2660 * 0: no write access, no denied write access 2661 * < 0: (-i_writecount) users that denied write access to the file. 2662 * > 0: (i_writecount) users that have write access to the file. 2663 * 2664 * Normally we operate on that counter with atomic_{inc,dec} and it's safe 2665 * except for the cases where we don't hold i_writecount yet. Then we need to 2666 * use {get,deny}_write_access() - these functions check the sign and refuse 2667 * to do the change if sign is wrong. 2668 */ 2669 static inline int get_write_access(struct inode *inode) 2670 { 2671 return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY; 2672 } 2673 static inline int deny_write_access(struct file *file) 2674 { 2675 struct inode *inode = file_inode(file); 2676 return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY; 2677 } 2678 static inline void put_write_access(struct inode * inode) 2679 { 2680 atomic_dec(&inode->i_writecount); 2681 } 2682 static inline void allow_write_access(struct file *file) 2683 { 2684 if (file) 2685 atomic_inc(&file_inode(file)->i_writecount); 2686 } 2687 static inline bool inode_is_open_for_write(const struct inode *inode) 2688 { 2689 return atomic_read(&inode->i_writecount) > 0; 2690 } 2691 2692 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING) 2693 static inline void i_readcount_dec(struct inode *inode) 2694 { 2695 BUG_ON(!atomic_read(&inode->i_readcount)); 2696 atomic_dec(&inode->i_readcount); 2697 } 2698 static inline void i_readcount_inc(struct inode *inode) 2699 { 2700 atomic_inc(&inode->i_readcount); 2701 } 2702 #else 2703 static inline void i_readcount_dec(struct inode *inode) 2704 { 2705 return; 2706 } 2707 static inline void i_readcount_inc(struct inode *inode) 2708 { 2709 return; 2710 } 2711 #endif 2712 extern int do_pipe_flags(int *, int); 2713 2714 extern ssize_t kernel_read(struct file *, void *, size_t, loff_t *); 2715 ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos); 2716 extern ssize_t kernel_write(struct file *, const void *, size_t, loff_t *); 2717 extern ssize_t __kernel_write(struct file *, const void *, size_t, loff_t *); 2718 extern struct file * open_exec(const char *); 2719 2720 /* fs/dcache.c -- generic fs support functions */ 2721 extern bool is_subdir(struct dentry *, struct dentry *); 2722 extern bool path_is_under(const struct path *, const struct path *); 2723 2724 extern char *file_path(struct file *, char *, int); 2725 2726 #include <linux/err.h> 2727 2728 /* needed for stackable file system support */ 2729 extern loff_t default_llseek(struct file *file, loff_t offset, int whence); 2730 2731 extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence); 2732 2733 extern int inode_init_always(struct super_block *, struct inode *); 2734 extern void inode_init_once(struct inode *); 2735 extern void address_space_init_once(struct address_space *mapping); 2736 extern struct inode * igrab(struct inode *); 2737 extern ino_t iunique(struct super_block *, ino_t); 2738 extern int inode_needs_sync(struct inode *inode); 2739 extern int generic_delete_inode(struct inode *inode); 2740 static inline int generic_drop_inode(struct inode *inode) 2741 { 2742 return !inode->i_nlink || inode_unhashed(inode); 2743 } 2744 extern void d_mark_dontcache(struct inode *inode); 2745 2746 extern struct inode *ilookup5_nowait(struct super_block *sb, 2747 unsigned long hashval, int (*test)(struct inode *, void *), 2748 void *data); 2749 extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval, 2750 int (*test)(struct inode *, void *), void *data); 2751 extern struct inode *ilookup(struct super_block *sb, unsigned long ino); 2752 2753 extern struct inode *inode_insert5(struct inode *inode, unsigned long hashval, 2754 int (*test)(struct inode *, void *), 2755 int (*set)(struct inode *, void *), 2756 void *data); 2757 extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *); 2758 extern struct inode * iget_locked(struct super_block *, unsigned long); 2759 extern struct inode *find_inode_nowait(struct super_block *, 2760 unsigned long, 2761 int (*match)(struct inode *, 2762 unsigned long, void *), 2763 void *data); 2764 extern struct inode *find_inode_rcu(struct super_block *, unsigned long, 2765 int (*)(struct inode *, void *), void *); 2766 extern struct inode *find_inode_by_ino_rcu(struct super_block *, unsigned long); 2767 extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *); 2768 extern int insert_inode_locked(struct inode *); 2769 #ifdef CONFIG_DEBUG_LOCK_ALLOC 2770 extern void lockdep_annotate_inode_mutex_key(struct inode *inode); 2771 #else 2772 static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { }; 2773 #endif 2774 extern void unlock_new_inode(struct inode *); 2775 extern void discard_new_inode(struct inode *); 2776 extern unsigned int get_next_ino(void); 2777 extern void evict_inodes(struct super_block *sb); 2778 void dump_mapping(const struct address_space *); 2779 2780 /* 2781 * Userspace may rely on the inode number being non-zero. For example, glibc 2782 * simply ignores files with zero i_ino in unlink() and other places. 2783 * 2784 * As an additional complication, if userspace was compiled with 2785 * _FILE_OFFSET_BITS=32 on a 64-bit kernel we'll only end up reading out the 2786 * lower 32 bits, so we need to check that those aren't zero explicitly. With 2787 * _FILE_OFFSET_BITS=64, this may cause some harmless false-negatives, but 2788 * better safe than sorry. 2789 */ 2790 static inline bool is_zero_ino(ino_t ino) 2791 { 2792 return (u32)ino == 0; 2793 } 2794 2795 extern void __iget(struct inode * inode); 2796 extern void iget_failed(struct inode *); 2797 extern void clear_inode(struct inode *); 2798 extern void __destroy_inode(struct inode *); 2799 extern struct inode *new_inode_pseudo(struct super_block *sb); 2800 extern struct inode *new_inode(struct super_block *sb); 2801 extern void free_inode_nonrcu(struct inode *inode); 2802 extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *); 2803 extern int file_remove_privs(struct file *); 2804 int setattr_should_drop_sgid(struct mnt_idmap *idmap, 2805 const struct inode *inode); 2806 2807 /* 2808 * This must be used for allocating filesystems specific inodes to set 2809 * up the inode reclaim context correctly. 2810 */ 2811 static inline void * 2812 alloc_inode_sb(struct super_block *sb, struct kmem_cache *cache, gfp_t gfp) 2813 { 2814 return kmem_cache_alloc_lru(cache, &sb->s_inode_lru, gfp); 2815 } 2816 2817 extern void __insert_inode_hash(struct inode *, unsigned long hashval); 2818 static inline void insert_inode_hash(struct inode *inode) 2819 { 2820 __insert_inode_hash(inode, inode->i_ino); 2821 } 2822 2823 extern void __remove_inode_hash(struct inode *); 2824 static inline void remove_inode_hash(struct inode *inode) 2825 { 2826 if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash)) 2827 __remove_inode_hash(inode); 2828 } 2829 2830 extern void inode_sb_list_add(struct inode *inode); 2831 extern void inode_add_lru(struct inode *inode); 2832 2833 extern int sb_set_blocksize(struct super_block *, int); 2834 extern int sb_min_blocksize(struct super_block *, int); 2835 2836 extern int generic_file_mmap(struct file *, struct vm_area_struct *); 2837 extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *); 2838 extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *); 2839 int generic_write_checks_count(struct kiocb *iocb, loff_t *count); 2840 extern int generic_write_check_limits(struct file *file, loff_t pos, 2841 loff_t *count); 2842 extern int generic_file_rw_checks(struct file *file_in, struct file *file_out); 2843 ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *to, 2844 ssize_t already_read); 2845 extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *); 2846 extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *); 2847 extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *); 2848 extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *); 2849 ssize_t generic_perform_write(struct kiocb *, struct iov_iter *); 2850 ssize_t direct_write_fallback(struct kiocb *iocb, struct iov_iter *iter, 2851 ssize_t direct_written, ssize_t buffered_written); 2852 2853 ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos, 2854 rwf_t flags); 2855 ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos, 2856 rwf_t flags); 2857 ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb, 2858 struct iov_iter *iter); 2859 ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb, 2860 struct iov_iter *iter); 2861 2862 /* fs/splice.c */ 2863 ssize_t filemap_splice_read(struct file *in, loff_t *ppos, 2864 struct pipe_inode_info *pipe, 2865 size_t len, unsigned int flags); 2866 ssize_t copy_splice_read(struct file *in, loff_t *ppos, 2867 struct pipe_inode_info *pipe, 2868 size_t len, unsigned int flags); 2869 extern ssize_t iter_file_splice_write(struct pipe_inode_info *, 2870 struct file *, loff_t *, size_t, unsigned int); 2871 extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out, 2872 loff_t *opos, size_t len, unsigned int flags); 2873 2874 2875 extern void 2876 file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping); 2877 extern loff_t noop_llseek(struct file *file, loff_t offset, int whence); 2878 #define no_llseek NULL 2879 extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize); 2880 extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence); 2881 extern loff_t generic_file_llseek_size(struct file *file, loff_t offset, 2882 int whence, loff_t maxsize, loff_t eof); 2883 extern loff_t fixed_size_llseek(struct file *file, loff_t offset, 2884 int whence, loff_t size); 2885 extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t); 2886 extern loff_t no_seek_end_llseek(struct file *, loff_t, int); 2887 int rw_verify_area(int, struct file *, const loff_t *, size_t); 2888 extern int generic_file_open(struct inode * inode, struct file * filp); 2889 extern int nonseekable_open(struct inode * inode, struct file * filp); 2890 extern int stream_open(struct inode * inode, struct file * filp); 2891 2892 #ifdef CONFIG_BLOCK 2893 typedef void (dio_submit_t)(struct bio *bio, struct inode *inode, 2894 loff_t file_offset); 2895 2896 enum { 2897 /* need locking between buffered and direct access */ 2898 DIO_LOCKING = 0x01, 2899 2900 /* filesystem does not support filling holes */ 2901 DIO_SKIP_HOLES = 0x02, 2902 }; 2903 2904 ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode, 2905 struct block_device *bdev, struct iov_iter *iter, 2906 get_block_t get_block, 2907 dio_iodone_t end_io, 2908 int flags); 2909 2910 static inline ssize_t blockdev_direct_IO(struct kiocb *iocb, 2911 struct inode *inode, 2912 struct iov_iter *iter, 2913 get_block_t get_block) 2914 { 2915 return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter, 2916 get_block, NULL, DIO_LOCKING | DIO_SKIP_HOLES); 2917 } 2918 #endif 2919 2920 void inode_dio_wait(struct inode *inode); 2921 2922 /** 2923 * inode_dio_begin - signal start of a direct I/O requests 2924 * @inode: inode the direct I/O happens on 2925 * 2926 * This is called once we've finished processing a direct I/O request, 2927 * and is used to wake up callers waiting for direct I/O to be quiesced. 2928 */ 2929 static inline void inode_dio_begin(struct inode *inode) 2930 { 2931 atomic_inc(&inode->i_dio_count); 2932 } 2933 2934 /** 2935 * inode_dio_end - signal finish of a direct I/O requests 2936 * @inode: inode the direct I/O happens on 2937 * 2938 * This is called once we've finished processing a direct I/O request, 2939 * and is used to wake up callers waiting for direct I/O to be quiesced. 2940 */ 2941 static inline void inode_dio_end(struct inode *inode) 2942 { 2943 if (atomic_dec_and_test(&inode->i_dio_count)) 2944 wake_up_bit(&inode->i_state, __I_DIO_WAKEUP); 2945 } 2946 2947 extern void inode_set_flags(struct inode *inode, unsigned int flags, 2948 unsigned int mask); 2949 2950 extern const struct file_operations generic_ro_fops; 2951 2952 #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m)) 2953 2954 extern int readlink_copy(char __user *, int, const char *); 2955 extern int page_readlink(struct dentry *, char __user *, int); 2956 extern const char *page_get_link(struct dentry *, struct inode *, 2957 struct delayed_call *); 2958 extern void page_put_link(void *); 2959 extern int page_symlink(struct inode *inode, const char *symname, int len); 2960 extern const struct inode_operations page_symlink_inode_operations; 2961 extern void kfree_link(void *); 2962 void fill_mg_cmtime(struct kstat *stat, u32 request_mask, struct inode *inode); 2963 void generic_fillattr(struct mnt_idmap *, u32, struct inode *, struct kstat *); 2964 void generic_fill_statx_attr(struct inode *inode, struct kstat *stat); 2965 extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int); 2966 extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int); 2967 void __inode_add_bytes(struct inode *inode, loff_t bytes); 2968 void inode_add_bytes(struct inode *inode, loff_t bytes); 2969 void __inode_sub_bytes(struct inode *inode, loff_t bytes); 2970 void inode_sub_bytes(struct inode *inode, loff_t bytes); 2971 static inline loff_t __inode_get_bytes(struct inode *inode) 2972 { 2973 return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes; 2974 } 2975 loff_t inode_get_bytes(struct inode *inode); 2976 void inode_set_bytes(struct inode *inode, loff_t bytes); 2977 const char *simple_get_link(struct dentry *, struct inode *, 2978 struct delayed_call *); 2979 extern const struct inode_operations simple_symlink_inode_operations; 2980 2981 extern int iterate_dir(struct file *, struct dir_context *); 2982 2983 int vfs_fstatat(int dfd, const char __user *filename, struct kstat *stat, 2984 int flags); 2985 int vfs_fstat(int fd, struct kstat *stat); 2986 2987 static inline int vfs_stat(const char __user *filename, struct kstat *stat) 2988 { 2989 return vfs_fstatat(AT_FDCWD, filename, stat, 0); 2990 } 2991 static inline int vfs_lstat(const char __user *name, struct kstat *stat) 2992 { 2993 return vfs_fstatat(AT_FDCWD, name, stat, AT_SYMLINK_NOFOLLOW); 2994 } 2995 2996 extern const char *vfs_get_link(struct dentry *, struct delayed_call *); 2997 extern int vfs_readlink(struct dentry *, char __user *, int); 2998 2999 extern struct file_system_type *get_filesystem(struct file_system_type *fs); 3000 extern void put_filesystem(struct file_system_type *fs); 3001 extern struct file_system_type *get_fs_type(const char *name); 3002 extern struct super_block *get_super(struct block_device *); 3003 extern struct super_block *get_active_super(struct block_device *bdev); 3004 extern void drop_super(struct super_block *sb); 3005 extern void drop_super_exclusive(struct super_block *sb); 3006 extern void iterate_supers(void (*)(struct super_block *, void *), void *); 3007 extern void iterate_supers_type(struct file_system_type *, 3008 void (*)(struct super_block *, void *), void *); 3009 3010 extern int dcache_dir_open(struct inode *, struct file *); 3011 extern int dcache_dir_close(struct inode *, struct file *); 3012 extern loff_t dcache_dir_lseek(struct file *, loff_t, int); 3013 extern int dcache_readdir(struct file *, struct dir_context *); 3014 extern int simple_setattr(struct mnt_idmap *, struct dentry *, 3015 struct iattr *); 3016 extern int simple_getattr(struct mnt_idmap *, const struct path *, 3017 struct kstat *, u32, unsigned int); 3018 extern int simple_statfs(struct dentry *, struct kstatfs *); 3019 extern int simple_open(struct inode *inode, struct file *file); 3020 extern int simple_link(struct dentry *, struct inode *, struct dentry *); 3021 extern int simple_unlink(struct inode *, struct dentry *); 3022 extern int simple_rmdir(struct inode *, struct dentry *); 3023 void simple_rename_timestamp(struct inode *old_dir, struct dentry *old_dentry, 3024 struct inode *new_dir, struct dentry *new_dentry); 3025 extern int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry, 3026 struct inode *new_dir, struct dentry *new_dentry); 3027 extern int simple_rename(struct mnt_idmap *, struct inode *, 3028 struct dentry *, struct inode *, struct dentry *, 3029 unsigned int); 3030 extern void simple_recursive_removal(struct dentry *, 3031 void (*callback)(struct dentry *)); 3032 extern int noop_fsync(struct file *, loff_t, loff_t, int); 3033 extern ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter); 3034 extern int simple_empty(struct dentry *); 3035 extern int simple_write_begin(struct file *file, struct address_space *mapping, 3036 loff_t pos, unsigned len, 3037 struct page **pagep, void **fsdata); 3038 extern const struct address_space_operations ram_aops; 3039 extern int always_delete_dentry(const struct dentry *); 3040 extern struct inode *alloc_anon_inode(struct super_block *); 3041 extern int simple_nosetlease(struct file *, long, struct file_lock **, void **); 3042 extern const struct dentry_operations simple_dentry_operations; 3043 3044 extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags); 3045 extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *); 3046 extern const struct file_operations simple_dir_operations; 3047 extern const struct inode_operations simple_dir_inode_operations; 3048 extern void make_empty_dir_inode(struct inode *inode); 3049 extern bool is_empty_dir_inode(struct inode *inode); 3050 struct tree_descr { const char *name; const struct file_operations *ops; int mode; }; 3051 struct dentry *d_alloc_name(struct dentry *, const char *); 3052 extern int simple_fill_super(struct super_block *, unsigned long, 3053 const struct tree_descr *); 3054 extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count); 3055 extern void simple_release_fs(struct vfsmount **mount, int *count); 3056 3057 extern ssize_t simple_read_from_buffer(void __user *to, size_t count, 3058 loff_t *ppos, const void *from, size_t available); 3059 extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos, 3060 const void __user *from, size_t count); 3061 3062 extern int __generic_file_fsync(struct file *, loff_t, loff_t, int); 3063 extern int generic_file_fsync(struct file *, loff_t, loff_t, int); 3064 3065 extern int generic_check_addressable(unsigned, u64); 3066 3067 extern void generic_set_encrypted_ci_d_ops(struct dentry *dentry); 3068 3069 int may_setattr(struct mnt_idmap *idmap, struct inode *inode, 3070 unsigned int ia_valid); 3071 int setattr_prepare(struct mnt_idmap *, struct dentry *, struct iattr *); 3072 extern int inode_newsize_ok(const struct inode *, loff_t offset); 3073 void setattr_copy(struct mnt_idmap *, struct inode *inode, 3074 const struct iattr *attr); 3075 3076 extern int file_update_time(struct file *file); 3077 3078 static inline bool vma_is_dax(const struct vm_area_struct *vma) 3079 { 3080 return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host); 3081 } 3082 3083 static inline bool vma_is_fsdax(struct vm_area_struct *vma) 3084 { 3085 struct inode *inode; 3086 3087 if (!IS_ENABLED(CONFIG_FS_DAX) || !vma->vm_file) 3088 return false; 3089 if (!vma_is_dax(vma)) 3090 return false; 3091 inode = file_inode(vma->vm_file); 3092 if (S_ISCHR(inode->i_mode)) 3093 return false; /* device-dax */ 3094 return true; 3095 } 3096 3097 static inline int iocb_flags(struct file *file) 3098 { 3099 int res = 0; 3100 if (file->f_flags & O_APPEND) 3101 res |= IOCB_APPEND; 3102 if (file->f_flags & O_DIRECT) 3103 res |= IOCB_DIRECT; 3104 if (file->f_flags & O_DSYNC) 3105 res |= IOCB_DSYNC; 3106 if (file->f_flags & __O_SYNC) 3107 res |= IOCB_SYNC; 3108 return res; 3109 } 3110 3111 static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags) 3112 { 3113 int kiocb_flags = 0; 3114 3115 /* make sure there's no overlap between RWF and private IOCB flags */ 3116 BUILD_BUG_ON((__force int) RWF_SUPPORTED & IOCB_EVENTFD); 3117 3118 if (!flags) 3119 return 0; 3120 if (unlikely(flags & ~RWF_SUPPORTED)) 3121 return -EOPNOTSUPP; 3122 3123 if (flags & RWF_NOWAIT) { 3124 if (!(ki->ki_filp->f_mode & FMODE_NOWAIT)) 3125 return -EOPNOTSUPP; 3126 kiocb_flags |= IOCB_NOIO; 3127 } 3128 kiocb_flags |= (__force int) (flags & RWF_SUPPORTED); 3129 if (flags & RWF_SYNC) 3130 kiocb_flags |= IOCB_DSYNC; 3131 3132 ki->ki_flags |= kiocb_flags; 3133 return 0; 3134 } 3135 3136 static inline ino_t parent_ino(struct dentry *dentry) 3137 { 3138 ino_t res; 3139 3140 /* 3141 * Don't strictly need d_lock here? If the parent ino could change 3142 * then surely we'd have a deeper race in the caller? 3143 */ 3144 spin_lock(&dentry->d_lock); 3145 res = dentry->d_parent->d_inode->i_ino; 3146 spin_unlock(&dentry->d_lock); 3147 return res; 3148 } 3149 3150 /* Transaction based IO helpers */ 3151 3152 /* 3153 * An argresp is stored in an allocated page and holds the 3154 * size of the argument or response, along with its content 3155 */ 3156 struct simple_transaction_argresp { 3157 ssize_t size; 3158 char data[]; 3159 }; 3160 3161 #define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp)) 3162 3163 char *simple_transaction_get(struct file *file, const char __user *buf, 3164 size_t size); 3165 ssize_t simple_transaction_read(struct file *file, char __user *buf, 3166 size_t size, loff_t *pos); 3167 int simple_transaction_release(struct inode *inode, struct file *file); 3168 3169 void simple_transaction_set(struct file *file, size_t n); 3170 3171 /* 3172 * simple attribute files 3173 * 3174 * These attributes behave similar to those in sysfs: 3175 * 3176 * Writing to an attribute immediately sets a value, an open file can be 3177 * written to multiple times. 3178 * 3179 * Reading from an attribute creates a buffer from the value that might get 3180 * read with multiple read calls. When the attribute has been read 3181 * completely, no further read calls are possible until the file is opened 3182 * again. 3183 * 3184 * All attributes contain a text representation of a numeric value 3185 * that are accessed with the get() and set() functions. 3186 */ 3187 #define DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, __is_signed) \ 3188 static int __fops ## _open(struct inode *inode, struct file *file) \ 3189 { \ 3190 __simple_attr_check_format(__fmt, 0ull); \ 3191 return simple_attr_open(inode, file, __get, __set, __fmt); \ 3192 } \ 3193 static const struct file_operations __fops = { \ 3194 .owner = THIS_MODULE, \ 3195 .open = __fops ## _open, \ 3196 .release = simple_attr_release, \ 3197 .read = simple_attr_read, \ 3198 .write = (__is_signed) ? simple_attr_write_signed : simple_attr_write, \ 3199 .llseek = generic_file_llseek, \ 3200 } 3201 3202 #define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt) \ 3203 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, false) 3204 3205 #define DEFINE_SIMPLE_ATTRIBUTE_SIGNED(__fops, __get, __set, __fmt) \ 3206 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, true) 3207 3208 static inline __printf(1, 2) 3209 void __simple_attr_check_format(const char *fmt, ...) 3210 { 3211 /* don't do anything, just let the compiler check the arguments; */ 3212 } 3213 3214 int simple_attr_open(struct inode *inode, struct file *file, 3215 int (*get)(void *, u64 *), int (*set)(void *, u64), 3216 const char *fmt); 3217 int simple_attr_release(struct inode *inode, struct file *file); 3218 ssize_t simple_attr_read(struct file *file, char __user *buf, 3219 size_t len, loff_t *ppos); 3220 ssize_t simple_attr_write(struct file *file, const char __user *buf, 3221 size_t len, loff_t *ppos); 3222 ssize_t simple_attr_write_signed(struct file *file, const char __user *buf, 3223 size_t len, loff_t *ppos); 3224 3225 struct ctl_table; 3226 int __init list_bdev_fs_names(char *buf, size_t size); 3227 3228 #define __FMODE_EXEC ((__force int) FMODE_EXEC) 3229 #define __FMODE_NONOTIFY ((__force int) FMODE_NONOTIFY) 3230 3231 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE]) 3232 #define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \ 3233 (flag & __FMODE_NONOTIFY))) 3234 3235 static inline bool is_sxid(umode_t mode) 3236 { 3237 return mode & (S_ISUID | S_ISGID); 3238 } 3239 3240 static inline int check_sticky(struct mnt_idmap *idmap, 3241 struct inode *dir, struct inode *inode) 3242 { 3243 if (!(dir->i_mode & S_ISVTX)) 3244 return 0; 3245 3246 return __check_sticky(idmap, dir, inode); 3247 } 3248 3249 static inline void inode_has_no_xattr(struct inode *inode) 3250 { 3251 if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC)) 3252 inode->i_flags |= S_NOSEC; 3253 } 3254 3255 static inline bool is_root_inode(struct inode *inode) 3256 { 3257 return inode == inode->i_sb->s_root->d_inode; 3258 } 3259 3260 static inline bool dir_emit(struct dir_context *ctx, 3261 const char *name, int namelen, 3262 u64 ino, unsigned type) 3263 { 3264 return ctx->actor(ctx, name, namelen, ctx->pos, ino, type); 3265 } 3266 static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx) 3267 { 3268 return ctx->actor(ctx, ".", 1, ctx->pos, 3269 file->f_path.dentry->d_inode->i_ino, DT_DIR); 3270 } 3271 static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx) 3272 { 3273 return ctx->actor(ctx, "..", 2, ctx->pos, 3274 parent_ino(file->f_path.dentry), DT_DIR); 3275 } 3276 static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx) 3277 { 3278 if (ctx->pos == 0) { 3279 if (!dir_emit_dot(file, ctx)) 3280 return false; 3281 ctx->pos = 1; 3282 } 3283 if (ctx->pos == 1) { 3284 if (!dir_emit_dotdot(file, ctx)) 3285 return false; 3286 ctx->pos = 2; 3287 } 3288 return true; 3289 } 3290 static inline bool dir_relax(struct inode *inode) 3291 { 3292 inode_unlock(inode); 3293 inode_lock(inode); 3294 return !IS_DEADDIR(inode); 3295 } 3296 3297 static inline bool dir_relax_shared(struct inode *inode) 3298 { 3299 inode_unlock_shared(inode); 3300 inode_lock_shared(inode); 3301 return !IS_DEADDIR(inode); 3302 } 3303 3304 extern bool path_noexec(const struct path *path); 3305 extern void inode_nohighmem(struct inode *inode); 3306 3307 /* mm/fadvise.c */ 3308 extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len, 3309 int advice); 3310 extern int generic_fadvise(struct file *file, loff_t offset, loff_t len, 3311 int advice); 3312 3313 #endif /* _LINUX_FS_H */ 3314