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 /* File is opened with O_NDELAY (only set for block devices) */ 123 #define FMODE_NDELAY ((__force fmode_t)0x40) 124 /* File is opened with O_EXCL (only set for block devices) */ 125 #define FMODE_EXCL ((__force fmode_t)0x80) 126 /* File is opened using open(.., 3, ..) and is writeable only for ioctls 127 (specialy hack for floppy.c) */ 128 #define FMODE_WRITE_IOCTL ((__force fmode_t)0x100) 129 /* 32bit hashes as llseek() offset (for directories) */ 130 #define FMODE_32BITHASH ((__force fmode_t)0x200) 131 /* 64bit hashes as llseek() offset (for directories) */ 132 #define FMODE_64BITHASH ((__force fmode_t)0x400) 133 134 /* 135 * Don't update ctime and mtime. 136 * 137 * Currently a special hack for the XFS open_by_handle ioctl, but we'll 138 * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon. 139 */ 140 #define FMODE_NOCMTIME ((__force fmode_t)0x800) 141 142 /* Expect random access pattern */ 143 #define FMODE_RANDOM ((__force fmode_t)0x1000) 144 145 /* File is huge (eg. /dev/mem): treat loff_t as unsigned */ 146 #define FMODE_UNSIGNED_OFFSET ((__force fmode_t)0x2000) 147 148 /* File is opened with O_PATH; almost nothing can be done with it */ 149 #define FMODE_PATH ((__force fmode_t)0x4000) 150 151 /* File needs atomic accesses to f_pos */ 152 #define FMODE_ATOMIC_POS ((__force fmode_t)0x8000) 153 /* Write access to underlying fs */ 154 #define FMODE_WRITER ((__force fmode_t)0x10000) 155 /* Has read method(s) */ 156 #define FMODE_CAN_READ ((__force fmode_t)0x20000) 157 /* Has write method(s) */ 158 #define FMODE_CAN_WRITE ((__force fmode_t)0x40000) 159 160 #define FMODE_OPENED ((__force fmode_t)0x80000) 161 #define FMODE_CREATED ((__force fmode_t)0x100000) 162 163 /* File is stream-like */ 164 #define FMODE_STREAM ((__force fmode_t)0x200000) 165 166 /* File supports DIRECT IO */ 167 #define FMODE_CAN_ODIRECT ((__force fmode_t)0x400000) 168 169 #define FMODE_NOREUSE ((__force fmode_t)0x800000) 170 171 /* File was opened by fanotify and shouldn't generate fanotify events */ 172 #define FMODE_NONOTIFY ((__force fmode_t)0x4000000) 173 174 /* File is capable of returning -EAGAIN if I/O will block */ 175 #define FMODE_NOWAIT ((__force fmode_t)0x8000000) 176 177 /* File represents mount that needs unmounting */ 178 #define FMODE_NEED_UNMOUNT ((__force fmode_t)0x10000000) 179 180 /* File does not contribute to nr_files count */ 181 #define FMODE_NOACCOUNT ((__force fmode_t)0x20000000) 182 183 /* File supports async buffered reads */ 184 #define FMODE_BUF_RASYNC ((__force fmode_t)0x40000000) 185 186 /* File supports async nowait buffered writes */ 187 #define FMODE_BUF_WASYNC ((__force fmode_t)0x80000000) 188 189 /* 190 * Attribute flags. These should be or-ed together to figure out what 191 * has been changed! 192 */ 193 #define ATTR_MODE (1 << 0) 194 #define ATTR_UID (1 << 1) 195 #define ATTR_GID (1 << 2) 196 #define ATTR_SIZE (1 << 3) 197 #define ATTR_ATIME (1 << 4) 198 #define ATTR_MTIME (1 << 5) 199 #define ATTR_CTIME (1 << 6) 200 #define ATTR_ATIME_SET (1 << 7) 201 #define ATTR_MTIME_SET (1 << 8) 202 #define ATTR_FORCE (1 << 9) /* Not a change, but a change it */ 203 #define ATTR_KILL_SUID (1 << 11) 204 #define ATTR_KILL_SGID (1 << 12) 205 #define ATTR_FILE (1 << 13) 206 #define ATTR_KILL_PRIV (1 << 14) 207 #define ATTR_OPEN (1 << 15) /* Truncating from open(O_TRUNC) */ 208 #define ATTR_TIMES_SET (1 << 16) 209 #define ATTR_TOUCH (1 << 17) 210 211 /* 212 * Whiteout is represented by a char device. The following constants define the 213 * mode and device number to use. 214 */ 215 #define WHITEOUT_MODE 0 216 #define WHITEOUT_DEV 0 217 218 /* 219 * This is the Inode Attributes structure, used for notify_change(). It 220 * uses the above definitions as flags, to know which values have changed. 221 * Also, in this manner, a Filesystem can look at only the values it cares 222 * about. Basically, these are the attributes that the VFS layer can 223 * request to change from the FS layer. 224 * 225 * Derek Atkins <warlord@MIT.EDU> 94-10-20 226 */ 227 struct iattr { 228 unsigned int ia_valid; 229 umode_t ia_mode; 230 /* 231 * The two anonymous unions wrap structures with the same member. 232 * 233 * Filesystems raising FS_ALLOW_IDMAP need to use ia_vfs{g,u}id which 234 * are a dedicated type requiring the filesystem to use the dedicated 235 * helpers. Other filesystem can continue to use ia_{g,u}id until they 236 * have been ported. 237 * 238 * They always contain the same value. In other words FS_ALLOW_IDMAP 239 * pass down the same value on idmapped mounts as they would on regular 240 * mounts. 241 */ 242 union { 243 kuid_t ia_uid; 244 vfsuid_t ia_vfsuid; 245 }; 246 union { 247 kgid_t ia_gid; 248 vfsgid_t ia_vfsgid; 249 }; 250 loff_t ia_size; 251 struct timespec64 ia_atime; 252 struct timespec64 ia_mtime; 253 struct timespec64 ia_ctime; 254 255 /* 256 * Not an attribute, but an auxiliary info for filesystems wanting to 257 * implement an ftruncate() like method. NOTE: filesystem should 258 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL). 259 */ 260 struct file *ia_file; 261 }; 262 263 /* 264 * Includes for diskquotas. 265 */ 266 #include <linux/quota.h> 267 268 /* 269 * Maximum number of layers of fs stack. Needs to be limited to 270 * prevent kernel stack overflow 271 */ 272 #define FILESYSTEM_MAX_STACK_DEPTH 2 273 274 /** 275 * enum positive_aop_returns - aop return codes with specific semantics 276 * 277 * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has 278 * completed, that the page is still locked, and 279 * should be considered active. The VM uses this hint 280 * to return the page to the active list -- it won't 281 * be a candidate for writeback again in the near 282 * future. Other callers must be careful to unlock 283 * the page if they get this return. Returned by 284 * writepage(); 285 * 286 * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has 287 * unlocked it and the page might have been truncated. 288 * The caller should back up to acquiring a new page and 289 * trying again. The aop will be taking reasonable 290 * precautions not to livelock. If the caller held a page 291 * reference, it should drop it before retrying. Returned 292 * by read_folio(). 293 * 294 * address_space_operation functions return these large constants to indicate 295 * special semantics to the caller. These are much larger than the bytes in a 296 * page to allow for functions that return the number of bytes operated on in a 297 * given page. 298 */ 299 300 enum positive_aop_returns { 301 AOP_WRITEPAGE_ACTIVATE = 0x80000, 302 AOP_TRUNCATED_PAGE = 0x80001, 303 }; 304 305 /* 306 * oh the beauties of C type declarations. 307 */ 308 struct page; 309 struct address_space; 310 struct writeback_control; 311 struct readahead_control; 312 313 /* 314 * Write life time hint values. 315 * Stored in struct inode as u8. 316 */ 317 enum rw_hint { 318 WRITE_LIFE_NOT_SET = 0, 319 WRITE_LIFE_NONE = RWH_WRITE_LIFE_NONE, 320 WRITE_LIFE_SHORT = RWH_WRITE_LIFE_SHORT, 321 WRITE_LIFE_MEDIUM = RWH_WRITE_LIFE_MEDIUM, 322 WRITE_LIFE_LONG = RWH_WRITE_LIFE_LONG, 323 WRITE_LIFE_EXTREME = RWH_WRITE_LIFE_EXTREME, 324 }; 325 326 /* Match RWF_* bits to IOCB bits */ 327 #define IOCB_HIPRI (__force int) RWF_HIPRI 328 #define IOCB_DSYNC (__force int) RWF_DSYNC 329 #define IOCB_SYNC (__force int) RWF_SYNC 330 #define IOCB_NOWAIT (__force int) RWF_NOWAIT 331 #define IOCB_APPEND (__force int) RWF_APPEND 332 333 /* non-RWF related bits - start at 16 */ 334 #define IOCB_EVENTFD (1 << 16) 335 #define IOCB_DIRECT (1 << 17) 336 #define IOCB_WRITE (1 << 18) 337 /* iocb->ki_waitq is valid */ 338 #define IOCB_WAITQ (1 << 19) 339 #define IOCB_NOIO (1 << 20) 340 /* can use bio alloc cache */ 341 #define IOCB_ALLOC_CACHE (1 << 21) 342 343 struct kiocb { 344 struct file *ki_filp; 345 loff_t ki_pos; 346 void (*ki_complete)(struct kiocb *iocb, long ret); 347 void *private; 348 int ki_flags; 349 u16 ki_ioprio; /* See linux/ioprio.h */ 350 struct wait_page_queue *ki_waitq; /* for async buffered IO */ 351 }; 352 353 static inline bool is_sync_kiocb(struct kiocb *kiocb) 354 { 355 return kiocb->ki_complete == NULL; 356 } 357 358 struct address_space_operations { 359 int (*writepage)(struct page *page, struct writeback_control *wbc); 360 int (*read_folio)(struct file *, struct folio *); 361 362 /* Write back some dirty pages from this mapping. */ 363 int (*writepages)(struct address_space *, struct writeback_control *); 364 365 /* Mark a folio dirty. Return true if this dirtied it */ 366 bool (*dirty_folio)(struct address_space *, struct folio *); 367 368 void (*readahead)(struct readahead_control *); 369 370 int (*write_begin)(struct file *, struct address_space *mapping, 371 loff_t pos, unsigned len, 372 struct page **pagep, void **fsdata); 373 int (*write_end)(struct file *, struct address_space *mapping, 374 loff_t pos, unsigned len, unsigned copied, 375 struct page *page, void *fsdata); 376 377 /* Unfortunately this kludge is needed for FIBMAP. Don't use it */ 378 sector_t (*bmap)(struct address_space *, sector_t); 379 void (*invalidate_folio) (struct folio *, size_t offset, size_t len); 380 bool (*release_folio)(struct folio *, gfp_t); 381 void (*free_folio)(struct folio *folio); 382 ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter); 383 /* 384 * migrate the contents of a folio to the specified target. If 385 * migrate_mode is MIGRATE_ASYNC, it must not block. 386 */ 387 int (*migrate_folio)(struct address_space *, struct folio *dst, 388 struct folio *src, enum migrate_mode); 389 int (*launder_folio)(struct folio *); 390 bool (*is_partially_uptodate) (struct folio *, size_t from, 391 size_t count); 392 void (*is_dirty_writeback) (struct folio *, bool *dirty, bool *wb); 393 int (*error_remove_page)(struct address_space *, struct page *); 394 395 /* swapfile support */ 396 int (*swap_activate)(struct swap_info_struct *sis, struct file *file, 397 sector_t *span); 398 void (*swap_deactivate)(struct file *file); 399 int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter); 400 }; 401 402 extern const struct address_space_operations empty_aops; 403 404 /** 405 * struct address_space - Contents of a cacheable, mappable object. 406 * @host: Owner, either the inode or the block_device. 407 * @i_pages: Cached pages. 408 * @invalidate_lock: Guards coherency between page cache contents and 409 * file offset->disk block mappings in the filesystem during invalidates. 410 * It is also used to block modification of page cache contents through 411 * memory mappings. 412 * @gfp_mask: Memory allocation flags to use for allocating pages. 413 * @i_mmap_writable: Number of VM_SHARED mappings. 414 * @nr_thps: Number of THPs in the pagecache (non-shmem only). 415 * @i_mmap: Tree of private and shared mappings. 416 * @i_mmap_rwsem: Protects @i_mmap and @i_mmap_writable. 417 * @nrpages: Number of page entries, protected by the i_pages lock. 418 * @writeback_index: Writeback starts here. 419 * @a_ops: Methods. 420 * @flags: Error bits and flags (AS_*). 421 * @wb_err: The most recent error which has occurred. 422 * @private_lock: For use by the owner of the address_space. 423 * @private_list: For use by the owner of the address_space. 424 * @private_data: For use by the owner of the address_space. 425 */ 426 struct address_space { 427 struct inode *host; 428 struct xarray i_pages; 429 struct rw_semaphore invalidate_lock; 430 gfp_t gfp_mask; 431 atomic_t i_mmap_writable; 432 #ifdef CONFIG_READ_ONLY_THP_FOR_FS 433 /* number of thp, only for non-shmem files */ 434 atomic_t nr_thps; 435 #endif 436 struct rb_root_cached i_mmap; 437 struct rw_semaphore i_mmap_rwsem; 438 unsigned long nrpages; 439 pgoff_t writeback_index; 440 const struct address_space_operations *a_ops; 441 unsigned long flags; 442 errseq_t wb_err; 443 spinlock_t private_lock; 444 struct list_head private_list; 445 void *private_data; 446 } __attribute__((aligned(sizeof(long)))) __randomize_layout; 447 /* 448 * On most architectures that alignment is already the case; but 449 * must be enforced here for CRIS, to let the least significant bit 450 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON. 451 */ 452 453 /* XArray tags, for tagging dirty and writeback pages in the pagecache. */ 454 #define PAGECACHE_TAG_DIRTY XA_MARK_0 455 #define PAGECACHE_TAG_WRITEBACK XA_MARK_1 456 #define PAGECACHE_TAG_TOWRITE XA_MARK_2 457 458 /* 459 * Returns true if any of the pages in the mapping are marked with the tag. 460 */ 461 static inline bool mapping_tagged(struct address_space *mapping, xa_mark_t tag) 462 { 463 return xa_marked(&mapping->i_pages, tag); 464 } 465 466 static inline void i_mmap_lock_write(struct address_space *mapping) 467 { 468 down_write(&mapping->i_mmap_rwsem); 469 } 470 471 static inline int i_mmap_trylock_write(struct address_space *mapping) 472 { 473 return down_write_trylock(&mapping->i_mmap_rwsem); 474 } 475 476 static inline void i_mmap_unlock_write(struct address_space *mapping) 477 { 478 up_write(&mapping->i_mmap_rwsem); 479 } 480 481 static inline int i_mmap_trylock_read(struct address_space *mapping) 482 { 483 return down_read_trylock(&mapping->i_mmap_rwsem); 484 } 485 486 static inline void i_mmap_lock_read(struct address_space *mapping) 487 { 488 down_read(&mapping->i_mmap_rwsem); 489 } 490 491 static inline void i_mmap_unlock_read(struct address_space *mapping) 492 { 493 up_read(&mapping->i_mmap_rwsem); 494 } 495 496 static inline void i_mmap_assert_locked(struct address_space *mapping) 497 { 498 lockdep_assert_held(&mapping->i_mmap_rwsem); 499 } 500 501 static inline void i_mmap_assert_write_locked(struct address_space *mapping) 502 { 503 lockdep_assert_held_write(&mapping->i_mmap_rwsem); 504 } 505 506 /* 507 * Might pages of this file be mapped into userspace? 508 */ 509 static inline int mapping_mapped(struct address_space *mapping) 510 { 511 return !RB_EMPTY_ROOT(&mapping->i_mmap.rb_root); 512 } 513 514 /* 515 * Might pages of this file have been modified in userspace? 516 * Note that i_mmap_writable counts all VM_SHARED vmas: do_mmap 517 * marks vma as VM_SHARED if it is shared, and the file was opened for 518 * writing i.e. vma may be mprotected writable even if now readonly. 519 * 520 * If i_mmap_writable is negative, no new writable mappings are allowed. You 521 * can only deny writable mappings, if none exists right now. 522 */ 523 static inline int mapping_writably_mapped(struct address_space *mapping) 524 { 525 return atomic_read(&mapping->i_mmap_writable) > 0; 526 } 527 528 static inline int mapping_map_writable(struct address_space *mapping) 529 { 530 return atomic_inc_unless_negative(&mapping->i_mmap_writable) ? 531 0 : -EPERM; 532 } 533 534 static inline void mapping_unmap_writable(struct address_space *mapping) 535 { 536 atomic_dec(&mapping->i_mmap_writable); 537 } 538 539 static inline int mapping_deny_writable(struct address_space *mapping) 540 { 541 return atomic_dec_unless_positive(&mapping->i_mmap_writable) ? 542 0 : -EBUSY; 543 } 544 545 static inline void mapping_allow_writable(struct address_space *mapping) 546 { 547 atomic_inc(&mapping->i_mmap_writable); 548 } 549 550 /* 551 * Use sequence counter to get consistent i_size on 32-bit processors. 552 */ 553 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 554 #include <linux/seqlock.h> 555 #define __NEED_I_SIZE_ORDERED 556 #define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount) 557 #else 558 #define i_size_ordered_init(inode) do { } while (0) 559 #endif 560 561 struct posix_acl; 562 #define ACL_NOT_CACHED ((void *)(-1)) 563 /* 564 * ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to 565 * cache the ACL. This also means that ->get_inode_acl() can be called in RCU 566 * mode with the LOOKUP_RCU flag. 567 */ 568 #define ACL_DONT_CACHE ((void *)(-3)) 569 570 static inline struct posix_acl * 571 uncached_acl_sentinel(struct task_struct *task) 572 { 573 return (void *)task + 1; 574 } 575 576 static inline bool 577 is_uncached_acl(struct posix_acl *acl) 578 { 579 return (long)acl & 1; 580 } 581 582 #define IOP_FASTPERM 0x0001 583 #define IOP_LOOKUP 0x0002 584 #define IOP_NOFOLLOW 0x0004 585 #define IOP_XATTR 0x0008 586 #define IOP_DEFAULT_READLINK 0x0010 587 588 struct fsnotify_mark_connector; 589 590 /* 591 * Keep mostly read-only and often accessed (especially for 592 * the RCU path lookup and 'stat' data) fields at the beginning 593 * of the 'struct inode' 594 */ 595 struct inode { 596 umode_t i_mode; 597 unsigned short i_opflags; 598 kuid_t i_uid; 599 kgid_t i_gid; 600 unsigned int i_flags; 601 602 #ifdef CONFIG_FS_POSIX_ACL 603 struct posix_acl *i_acl; 604 struct posix_acl *i_default_acl; 605 #endif 606 607 const struct inode_operations *i_op; 608 struct super_block *i_sb; 609 struct address_space *i_mapping; 610 611 #ifdef CONFIG_SECURITY 612 void *i_security; 613 #endif 614 615 /* Stat data, not accessed from path walking */ 616 unsigned long i_ino; 617 /* 618 * Filesystems may only read i_nlink directly. They shall use the 619 * following functions for modification: 620 * 621 * (set|clear|inc|drop)_nlink 622 * inode_(inc|dec)_link_count 623 */ 624 union { 625 const unsigned int i_nlink; 626 unsigned int __i_nlink; 627 }; 628 dev_t i_rdev; 629 loff_t i_size; 630 struct timespec64 i_atime; 631 struct timespec64 i_mtime; 632 struct timespec64 i_ctime; 633 spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */ 634 unsigned short i_bytes; 635 u8 i_blkbits; 636 u8 i_write_hint; 637 blkcnt_t i_blocks; 638 639 #ifdef __NEED_I_SIZE_ORDERED 640 seqcount_t i_size_seqcount; 641 #endif 642 643 /* Misc */ 644 unsigned long i_state; 645 struct rw_semaphore i_rwsem; 646 647 unsigned long dirtied_when; /* jiffies of first dirtying */ 648 unsigned long dirtied_time_when; 649 650 struct hlist_node i_hash; 651 struct list_head i_io_list; /* backing dev IO list */ 652 #ifdef CONFIG_CGROUP_WRITEBACK 653 struct bdi_writeback *i_wb; /* the associated cgroup wb */ 654 655 /* foreign inode detection, see wbc_detach_inode() */ 656 int i_wb_frn_winner; 657 u16 i_wb_frn_avg_time; 658 u16 i_wb_frn_history; 659 #endif 660 struct list_head i_lru; /* inode LRU list */ 661 struct list_head i_sb_list; 662 struct list_head i_wb_list; /* backing dev writeback list */ 663 union { 664 struct hlist_head i_dentry; 665 struct rcu_head i_rcu; 666 }; 667 atomic64_t i_version; 668 atomic64_t i_sequence; /* see futex */ 669 atomic_t i_count; 670 atomic_t i_dio_count; 671 atomic_t i_writecount; 672 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING) 673 atomic_t i_readcount; /* struct files open RO */ 674 #endif 675 union { 676 const struct file_operations *i_fop; /* former ->i_op->default_file_ops */ 677 void (*free_inode)(struct inode *); 678 }; 679 struct file_lock_context *i_flctx; 680 struct address_space i_data; 681 struct list_head i_devices; 682 union { 683 struct pipe_inode_info *i_pipe; 684 struct cdev *i_cdev; 685 char *i_link; 686 unsigned i_dir_seq; 687 }; 688 689 __u32 i_generation; 690 691 #ifdef CONFIG_FSNOTIFY 692 __u32 i_fsnotify_mask; /* all events this inode cares about */ 693 struct fsnotify_mark_connector __rcu *i_fsnotify_marks; 694 #endif 695 696 #ifdef CONFIG_FS_ENCRYPTION 697 struct fscrypt_info *i_crypt_info; 698 #endif 699 700 #ifdef CONFIG_FS_VERITY 701 struct fsverity_info *i_verity_info; 702 #endif 703 704 void *i_private; /* fs or device private pointer */ 705 } __randomize_layout; 706 707 struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode); 708 709 static inline unsigned int i_blocksize(const struct inode *node) 710 { 711 return (1 << node->i_blkbits); 712 } 713 714 static inline int inode_unhashed(struct inode *inode) 715 { 716 return hlist_unhashed(&inode->i_hash); 717 } 718 719 /* 720 * __mark_inode_dirty expects inodes to be hashed. Since we don't 721 * want special inodes in the fileset inode space, we make them 722 * appear hashed, but do not put on any lists. hlist_del() 723 * will work fine and require no locking. 724 */ 725 static inline void inode_fake_hash(struct inode *inode) 726 { 727 hlist_add_fake(&inode->i_hash); 728 } 729 730 /* 731 * inode->i_mutex nesting subclasses for the lock validator: 732 * 733 * 0: the object of the current VFS operation 734 * 1: parent 735 * 2: child/target 736 * 3: xattr 737 * 4: second non-directory 738 * 5: second parent (when locking independent directories in rename) 739 * 740 * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two 741 * non-directories at once. 742 * 743 * The locking order between these classes is 744 * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory 745 */ 746 enum inode_i_mutex_lock_class 747 { 748 I_MUTEX_NORMAL, 749 I_MUTEX_PARENT, 750 I_MUTEX_CHILD, 751 I_MUTEX_XATTR, 752 I_MUTEX_NONDIR2, 753 I_MUTEX_PARENT2, 754 }; 755 756 static inline void inode_lock(struct inode *inode) 757 { 758 down_write(&inode->i_rwsem); 759 } 760 761 static inline void inode_unlock(struct inode *inode) 762 { 763 up_write(&inode->i_rwsem); 764 } 765 766 static inline void inode_lock_shared(struct inode *inode) 767 { 768 down_read(&inode->i_rwsem); 769 } 770 771 static inline void inode_unlock_shared(struct inode *inode) 772 { 773 up_read(&inode->i_rwsem); 774 } 775 776 static inline int inode_trylock(struct inode *inode) 777 { 778 return down_write_trylock(&inode->i_rwsem); 779 } 780 781 static inline int inode_trylock_shared(struct inode *inode) 782 { 783 return down_read_trylock(&inode->i_rwsem); 784 } 785 786 static inline int inode_is_locked(struct inode *inode) 787 { 788 return rwsem_is_locked(&inode->i_rwsem); 789 } 790 791 static inline void inode_lock_nested(struct inode *inode, unsigned subclass) 792 { 793 down_write_nested(&inode->i_rwsem, subclass); 794 } 795 796 static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass) 797 { 798 down_read_nested(&inode->i_rwsem, subclass); 799 } 800 801 static inline void filemap_invalidate_lock(struct address_space *mapping) 802 { 803 down_write(&mapping->invalidate_lock); 804 } 805 806 static inline void filemap_invalidate_unlock(struct address_space *mapping) 807 { 808 up_write(&mapping->invalidate_lock); 809 } 810 811 static inline void filemap_invalidate_lock_shared(struct address_space *mapping) 812 { 813 down_read(&mapping->invalidate_lock); 814 } 815 816 static inline int filemap_invalidate_trylock_shared( 817 struct address_space *mapping) 818 { 819 return down_read_trylock(&mapping->invalidate_lock); 820 } 821 822 static inline void filemap_invalidate_unlock_shared( 823 struct address_space *mapping) 824 { 825 up_read(&mapping->invalidate_lock); 826 } 827 828 void lock_two_nondirectories(struct inode *, struct inode*); 829 void unlock_two_nondirectories(struct inode *, struct inode*); 830 831 void filemap_invalidate_lock_two(struct address_space *mapping1, 832 struct address_space *mapping2); 833 void filemap_invalidate_unlock_two(struct address_space *mapping1, 834 struct address_space *mapping2); 835 836 837 /* 838 * NOTE: in a 32bit arch with a preemptable kernel and 839 * an UP compile the i_size_read/write must be atomic 840 * with respect to the local cpu (unlike with preempt disabled), 841 * but they don't need to be atomic with respect to other cpus like in 842 * true SMP (so they need either to either locally disable irq around 843 * the read or for example on x86 they can be still implemented as a 844 * cmpxchg8b without the need of the lock prefix). For SMP compiles 845 * and 64bit archs it makes no difference if preempt is enabled or not. 846 */ 847 static inline loff_t i_size_read(const struct inode *inode) 848 { 849 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 850 loff_t i_size; 851 unsigned int seq; 852 853 do { 854 seq = read_seqcount_begin(&inode->i_size_seqcount); 855 i_size = inode->i_size; 856 } while (read_seqcount_retry(&inode->i_size_seqcount, seq)); 857 return i_size; 858 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION) 859 loff_t i_size; 860 861 preempt_disable(); 862 i_size = inode->i_size; 863 preempt_enable(); 864 return i_size; 865 #else 866 return inode->i_size; 867 #endif 868 } 869 870 /* 871 * NOTE: unlike i_size_read(), i_size_write() does need locking around it 872 * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount 873 * can be lost, resulting in subsequent i_size_read() calls spinning forever. 874 */ 875 static inline void i_size_write(struct inode *inode, loff_t i_size) 876 { 877 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 878 preempt_disable(); 879 write_seqcount_begin(&inode->i_size_seqcount); 880 inode->i_size = i_size; 881 write_seqcount_end(&inode->i_size_seqcount); 882 preempt_enable(); 883 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION) 884 preempt_disable(); 885 inode->i_size = i_size; 886 preempt_enable(); 887 #else 888 inode->i_size = i_size; 889 #endif 890 } 891 892 static inline unsigned iminor(const struct inode *inode) 893 { 894 return MINOR(inode->i_rdev); 895 } 896 897 static inline unsigned imajor(const struct inode *inode) 898 { 899 return MAJOR(inode->i_rdev); 900 } 901 902 struct fown_struct { 903 rwlock_t lock; /* protects pid, uid, euid fields */ 904 struct pid *pid; /* pid or -pgrp where SIGIO should be sent */ 905 enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */ 906 kuid_t uid, euid; /* uid/euid of process setting the owner */ 907 int signum; /* posix.1b rt signal to be delivered on IO */ 908 }; 909 910 /** 911 * struct file_ra_state - Track a file's readahead state. 912 * @start: Where the most recent readahead started. 913 * @size: Number of pages read in the most recent readahead. 914 * @async_size: Numer of pages that were/are not needed immediately 915 * and so were/are genuinely "ahead". Start next readahead when 916 * the first of these pages is accessed. 917 * @ra_pages: Maximum size of a readahead request, copied from the bdi. 918 * @mmap_miss: How many mmap accesses missed in the page cache. 919 * @prev_pos: The last byte in the most recent read request. 920 * 921 * When this structure is passed to ->readahead(), the "most recent" 922 * readahead means the current readahead. 923 */ 924 struct file_ra_state { 925 pgoff_t start; 926 unsigned int size; 927 unsigned int async_size; 928 unsigned int ra_pages; 929 unsigned int mmap_miss; 930 loff_t prev_pos; 931 }; 932 933 /* 934 * Check if @index falls in the readahead windows. 935 */ 936 static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index) 937 { 938 return (index >= ra->start && 939 index < ra->start + ra->size); 940 } 941 942 struct file { 943 union { 944 struct llist_node f_llist; 945 struct rcu_head f_rcuhead; 946 unsigned int f_iocb_flags; 947 }; 948 struct path f_path; 949 struct inode *f_inode; /* cached value */ 950 const struct file_operations *f_op; 951 952 /* 953 * Protects f_ep, f_flags. 954 * Must not be taken from IRQ context. 955 */ 956 spinlock_t f_lock; 957 atomic_long_t f_count; 958 unsigned int f_flags; 959 fmode_t f_mode; 960 struct mutex f_pos_lock; 961 loff_t f_pos; 962 struct fown_struct f_owner; 963 const struct cred *f_cred; 964 struct file_ra_state f_ra; 965 966 u64 f_version; 967 #ifdef CONFIG_SECURITY 968 void *f_security; 969 #endif 970 /* needed for tty driver, and maybe others */ 971 void *private_data; 972 973 #ifdef CONFIG_EPOLL 974 /* Used by fs/eventpoll.c to link all the hooks to this file */ 975 struct hlist_head *f_ep; 976 #endif /* #ifdef CONFIG_EPOLL */ 977 struct address_space *f_mapping; 978 errseq_t f_wb_err; 979 errseq_t f_sb_err; /* for syncfs */ 980 } __randomize_layout 981 __attribute__((aligned(4))); /* lest something weird decides that 2 is OK */ 982 983 struct file_handle { 984 __u32 handle_bytes; 985 int handle_type; 986 /* file identifier */ 987 unsigned char f_handle[]; 988 }; 989 990 static inline struct file *get_file(struct file *f) 991 { 992 atomic_long_inc(&f->f_count); 993 return f; 994 } 995 #define get_file_rcu(x) atomic_long_inc_not_zero(&(x)->f_count) 996 #define file_count(x) atomic_long_read(&(x)->f_count) 997 998 #define MAX_NON_LFS ((1UL<<31) - 1) 999 1000 /* Page cache limit. The filesystems should put that into their s_maxbytes 1001 limits, otherwise bad things can happen in VM. */ 1002 #if BITS_PER_LONG==32 1003 #define MAX_LFS_FILESIZE ((loff_t)ULONG_MAX << PAGE_SHIFT) 1004 #elif BITS_PER_LONG==64 1005 #define MAX_LFS_FILESIZE ((loff_t)LLONG_MAX) 1006 #endif 1007 1008 /* legacy typedef, should eventually be removed */ 1009 typedef void *fl_owner_t; 1010 1011 struct file_lock; 1012 1013 /* The following constant reflects the upper bound of the file/locking space */ 1014 #ifndef OFFSET_MAX 1015 #define OFFSET_MAX type_max(loff_t) 1016 #define OFFT_OFFSET_MAX type_max(off_t) 1017 #endif 1018 1019 extern void send_sigio(struct fown_struct *fown, int fd, int band); 1020 1021 static inline struct inode *file_inode(const struct file *f) 1022 { 1023 return f->f_inode; 1024 } 1025 1026 static inline struct dentry *file_dentry(const struct file *file) 1027 { 1028 return d_real(file->f_path.dentry, file_inode(file)); 1029 } 1030 1031 struct fasync_struct { 1032 rwlock_t fa_lock; 1033 int magic; 1034 int fa_fd; 1035 struct fasync_struct *fa_next; /* singly linked list */ 1036 struct file *fa_file; 1037 struct rcu_head fa_rcu; 1038 }; 1039 1040 #define FASYNC_MAGIC 0x4601 1041 1042 /* SMP safe fasync helpers: */ 1043 extern int fasync_helper(int, struct file *, int, struct fasync_struct **); 1044 extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *); 1045 extern int fasync_remove_entry(struct file *, struct fasync_struct **); 1046 extern struct fasync_struct *fasync_alloc(void); 1047 extern void fasync_free(struct fasync_struct *); 1048 1049 /* can be called from interrupts */ 1050 extern void kill_fasync(struct fasync_struct **, int, int); 1051 1052 extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force); 1053 extern int f_setown(struct file *filp, unsigned long arg, int force); 1054 extern void f_delown(struct file *filp); 1055 extern pid_t f_getown(struct file *filp); 1056 extern int send_sigurg(struct fown_struct *fown); 1057 1058 /* 1059 * sb->s_flags. Note that these mirror the equivalent MS_* flags where 1060 * represented in both. 1061 */ 1062 #define SB_RDONLY 1 /* Mount read-only */ 1063 #define SB_NOSUID 2 /* Ignore suid and sgid bits */ 1064 #define SB_NODEV 4 /* Disallow access to device special files */ 1065 #define SB_NOEXEC 8 /* Disallow program execution */ 1066 #define SB_SYNCHRONOUS 16 /* Writes are synced at once */ 1067 #define SB_MANDLOCK 64 /* Allow mandatory locks on an FS */ 1068 #define SB_DIRSYNC 128 /* Directory modifications are synchronous */ 1069 #define SB_NOATIME 1024 /* Do not update access times. */ 1070 #define SB_NODIRATIME 2048 /* Do not update directory access times */ 1071 #define SB_SILENT 32768 1072 #define SB_POSIXACL (1<<16) /* VFS does not apply the umask */ 1073 #define SB_INLINECRYPT (1<<17) /* Use blk-crypto for encrypted files */ 1074 #define SB_KERNMOUNT (1<<22) /* this is a kern_mount call */ 1075 #define SB_I_VERSION (1<<23) /* Update inode I_version field */ 1076 #define SB_LAZYTIME (1<<25) /* Update the on-disk [acm]times lazily */ 1077 1078 /* These sb flags are internal to the kernel */ 1079 #define SB_SUBMOUNT (1<<26) 1080 #define SB_FORCE (1<<27) 1081 #define SB_NOSEC (1<<28) 1082 #define SB_BORN (1<<29) 1083 #define SB_ACTIVE (1<<30) 1084 #define SB_NOUSER (1<<31) 1085 1086 /* These flags relate to encoding and casefolding */ 1087 #define SB_ENC_STRICT_MODE_FL (1 << 0) 1088 1089 #define sb_has_strict_encoding(sb) \ 1090 (sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL) 1091 1092 /* 1093 * Umount options 1094 */ 1095 1096 #define MNT_FORCE 0x00000001 /* Attempt to forcibily umount */ 1097 #define MNT_DETACH 0x00000002 /* Just detach from the tree */ 1098 #define MNT_EXPIRE 0x00000004 /* Mark for expiry */ 1099 #define UMOUNT_NOFOLLOW 0x00000008 /* Don't follow symlink on umount */ 1100 #define UMOUNT_UNUSED 0x80000000 /* Flag guaranteed to be unused */ 1101 1102 /* sb->s_iflags */ 1103 #define SB_I_CGROUPWB 0x00000001 /* cgroup-aware writeback enabled */ 1104 #define SB_I_NOEXEC 0x00000002 /* Ignore executables on this fs */ 1105 #define SB_I_NODEV 0x00000004 /* Ignore devices on this fs */ 1106 #define SB_I_STABLE_WRITES 0x00000008 /* don't modify blks until WB is done */ 1107 1108 /* sb->s_iflags to limit user namespace mounts */ 1109 #define SB_I_USERNS_VISIBLE 0x00000010 /* fstype already mounted */ 1110 #define SB_I_IMA_UNVERIFIABLE_SIGNATURE 0x00000020 1111 #define SB_I_UNTRUSTED_MOUNTER 0x00000040 1112 1113 #define SB_I_SKIP_SYNC 0x00000100 /* Skip superblock at global sync */ 1114 #define SB_I_PERSB_BDI 0x00000200 /* has a per-sb bdi */ 1115 #define SB_I_TS_EXPIRY_WARNED 0x00000400 /* warned about timestamp range expiry */ 1116 #define SB_I_RETIRED 0x00000800 /* superblock shouldn't be reused */ 1117 1118 /* Possible states of 'frozen' field */ 1119 enum { 1120 SB_UNFROZEN = 0, /* FS is unfrozen */ 1121 SB_FREEZE_WRITE = 1, /* Writes, dir ops, ioctls frozen */ 1122 SB_FREEZE_PAGEFAULT = 2, /* Page faults stopped as well */ 1123 SB_FREEZE_FS = 3, /* For internal FS use (e.g. to stop 1124 * internal threads if needed) */ 1125 SB_FREEZE_COMPLETE = 4, /* ->freeze_fs finished successfully */ 1126 }; 1127 1128 #define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1) 1129 1130 struct sb_writers { 1131 int frozen; /* Is sb frozen? */ 1132 wait_queue_head_t wait_unfrozen; /* wait for thaw */ 1133 struct percpu_rw_semaphore rw_sem[SB_FREEZE_LEVELS]; 1134 }; 1135 1136 struct super_block { 1137 struct list_head s_list; /* Keep this first */ 1138 dev_t s_dev; /* search index; _not_ kdev_t */ 1139 unsigned char s_blocksize_bits; 1140 unsigned long s_blocksize; 1141 loff_t s_maxbytes; /* Max file size */ 1142 struct file_system_type *s_type; 1143 const struct super_operations *s_op; 1144 const struct dquot_operations *dq_op; 1145 const struct quotactl_ops *s_qcop; 1146 const struct export_operations *s_export_op; 1147 unsigned long s_flags; 1148 unsigned long s_iflags; /* internal SB_I_* flags */ 1149 unsigned long s_magic; 1150 struct dentry *s_root; 1151 struct rw_semaphore s_umount; 1152 int s_count; 1153 atomic_t s_active; 1154 #ifdef CONFIG_SECURITY 1155 void *s_security; 1156 #endif 1157 const struct xattr_handler **s_xattr; 1158 #ifdef CONFIG_FS_ENCRYPTION 1159 const struct fscrypt_operations *s_cop; 1160 struct fscrypt_keyring *s_master_keys; /* master crypto keys in use */ 1161 #endif 1162 #ifdef CONFIG_FS_VERITY 1163 const struct fsverity_operations *s_vop; 1164 #endif 1165 #if IS_ENABLED(CONFIG_UNICODE) 1166 struct unicode_map *s_encoding; 1167 __u16 s_encoding_flags; 1168 #endif 1169 struct hlist_bl_head s_roots; /* alternate root dentries for NFS */ 1170 struct list_head s_mounts; /* list of mounts; _not_ for fs use */ 1171 struct block_device *s_bdev; 1172 struct backing_dev_info *s_bdi; 1173 struct mtd_info *s_mtd; 1174 struct hlist_node s_instances; 1175 unsigned int s_quota_types; /* Bitmask of supported quota types */ 1176 struct quota_info s_dquot; /* Diskquota specific options */ 1177 1178 struct sb_writers s_writers; 1179 1180 /* 1181 * Keep s_fs_info, s_time_gran, s_fsnotify_mask, and 1182 * s_fsnotify_marks together for cache efficiency. They are frequently 1183 * accessed and rarely modified. 1184 */ 1185 void *s_fs_info; /* Filesystem private info */ 1186 1187 /* Granularity of c/m/atime in ns (cannot be worse than a second) */ 1188 u32 s_time_gran; 1189 /* Time limits for c/m/atime in seconds */ 1190 time64_t s_time_min; 1191 time64_t s_time_max; 1192 #ifdef CONFIG_FSNOTIFY 1193 __u32 s_fsnotify_mask; 1194 struct fsnotify_mark_connector __rcu *s_fsnotify_marks; 1195 #endif 1196 1197 char s_id[32]; /* Informational name */ 1198 uuid_t s_uuid; /* UUID */ 1199 1200 unsigned int s_max_links; 1201 fmode_t s_mode; 1202 1203 /* 1204 * The next field is for VFS *only*. No filesystems have any business 1205 * even looking at it. You had been warned. 1206 */ 1207 struct mutex s_vfs_rename_mutex; /* Kludge */ 1208 1209 /* 1210 * Filesystem subtype. If non-empty the filesystem type field 1211 * in /proc/mounts will be "type.subtype" 1212 */ 1213 const char *s_subtype; 1214 1215 const struct dentry_operations *s_d_op; /* default d_op for dentries */ 1216 1217 struct shrinker s_shrink; /* per-sb shrinker handle */ 1218 1219 /* Number of inodes with nlink == 0 but still referenced */ 1220 atomic_long_t s_remove_count; 1221 1222 /* 1223 * Number of inode/mount/sb objects that are being watched, note that 1224 * inodes objects are currently double-accounted. 1225 */ 1226 atomic_long_t s_fsnotify_connectors; 1227 1228 /* Being remounted read-only */ 1229 int s_readonly_remount; 1230 1231 /* per-sb errseq_t for reporting writeback errors via syncfs */ 1232 errseq_t s_wb_err; 1233 1234 /* AIO completions deferred from interrupt context */ 1235 struct workqueue_struct *s_dio_done_wq; 1236 struct hlist_head s_pins; 1237 1238 /* 1239 * Owning user namespace and default context in which to 1240 * interpret filesystem uids, gids, quotas, device nodes, 1241 * xattrs and security labels. 1242 */ 1243 struct user_namespace *s_user_ns; 1244 1245 /* 1246 * The list_lru structure is essentially just a pointer to a table 1247 * of per-node lru lists, each of which has its own spinlock. 1248 * There is no need to put them into separate cachelines. 1249 */ 1250 struct list_lru s_dentry_lru; 1251 struct list_lru s_inode_lru; 1252 struct rcu_head rcu; 1253 struct work_struct destroy_work; 1254 1255 struct mutex s_sync_lock; /* sync serialisation lock */ 1256 1257 /* 1258 * Indicates how deep in a filesystem stack this SB is 1259 */ 1260 int s_stack_depth; 1261 1262 /* s_inode_list_lock protects s_inodes */ 1263 spinlock_t s_inode_list_lock ____cacheline_aligned_in_smp; 1264 struct list_head s_inodes; /* all inodes */ 1265 1266 spinlock_t s_inode_wblist_lock; 1267 struct list_head s_inodes_wb; /* writeback inodes */ 1268 } __randomize_layout; 1269 1270 static inline struct user_namespace *i_user_ns(const struct inode *inode) 1271 { 1272 return inode->i_sb->s_user_ns; 1273 } 1274 1275 /* Helper functions so that in most cases filesystems will 1276 * not need to deal directly with kuid_t and kgid_t and can 1277 * instead deal with the raw numeric values that are stored 1278 * in the filesystem. 1279 */ 1280 static inline uid_t i_uid_read(const struct inode *inode) 1281 { 1282 return from_kuid(i_user_ns(inode), inode->i_uid); 1283 } 1284 1285 static inline gid_t i_gid_read(const struct inode *inode) 1286 { 1287 return from_kgid(i_user_ns(inode), inode->i_gid); 1288 } 1289 1290 static inline void i_uid_write(struct inode *inode, uid_t uid) 1291 { 1292 inode->i_uid = make_kuid(i_user_ns(inode), uid); 1293 } 1294 1295 static inline void i_gid_write(struct inode *inode, gid_t gid) 1296 { 1297 inode->i_gid = make_kgid(i_user_ns(inode), gid); 1298 } 1299 1300 /** 1301 * i_uid_into_vfsuid - map an inode's i_uid down according to an idmapping 1302 * @idmap: idmap of the mount the inode was found from 1303 * @inode: inode to map 1304 * 1305 * Return: whe inode's i_uid mapped down according to @idmap. 1306 * If the inode's i_uid has no mapping INVALID_VFSUID is returned. 1307 */ 1308 static inline vfsuid_t i_uid_into_vfsuid(struct mnt_idmap *idmap, 1309 const struct inode *inode) 1310 { 1311 return make_vfsuid(idmap, i_user_ns(inode), inode->i_uid); 1312 } 1313 1314 /** 1315 * i_uid_needs_update - check whether inode's i_uid needs to be updated 1316 * @idmap: idmap of the mount the inode was found from 1317 * @attr: the new attributes of @inode 1318 * @inode: the inode to update 1319 * 1320 * Check whether the $inode's i_uid field needs to be updated taking idmapped 1321 * mounts into account if the filesystem supports it. 1322 * 1323 * Return: true if @inode's i_uid field needs to be updated, false if not. 1324 */ 1325 static inline bool i_uid_needs_update(struct mnt_idmap *idmap, 1326 const struct iattr *attr, 1327 const struct inode *inode) 1328 { 1329 return ((attr->ia_valid & ATTR_UID) && 1330 !vfsuid_eq(attr->ia_vfsuid, 1331 i_uid_into_vfsuid(idmap, inode))); 1332 } 1333 1334 /** 1335 * i_uid_update - update @inode's i_uid field 1336 * @idmap: idmap of the mount the inode was found from 1337 * @attr: the new attributes of @inode 1338 * @inode: the inode to update 1339 * 1340 * Safely update @inode's i_uid field translating the vfsuid of any idmapped 1341 * mount into the filesystem kuid. 1342 */ 1343 static inline void i_uid_update(struct mnt_idmap *idmap, 1344 const struct iattr *attr, 1345 struct inode *inode) 1346 { 1347 if (attr->ia_valid & ATTR_UID) 1348 inode->i_uid = from_vfsuid(idmap, i_user_ns(inode), 1349 attr->ia_vfsuid); 1350 } 1351 1352 /** 1353 * i_gid_into_vfsgid - map an inode's i_gid down according to an idmapping 1354 * @idmap: idmap of the mount the inode was found from 1355 * @inode: inode to map 1356 * 1357 * Return: the inode's i_gid mapped down according to @idmap. 1358 * If the inode's i_gid has no mapping INVALID_VFSGID is returned. 1359 */ 1360 static inline vfsgid_t i_gid_into_vfsgid(struct mnt_idmap *idmap, 1361 const struct inode *inode) 1362 { 1363 return make_vfsgid(idmap, i_user_ns(inode), inode->i_gid); 1364 } 1365 1366 /** 1367 * i_gid_needs_update - check whether inode's i_gid needs to be updated 1368 * @idmap: idmap of the mount the inode was found from 1369 * @attr: the new attributes of @inode 1370 * @inode: the inode to update 1371 * 1372 * Check whether the $inode's i_gid field needs to be updated taking idmapped 1373 * mounts into account if the filesystem supports it. 1374 * 1375 * Return: true if @inode's i_gid field needs to be updated, false if not. 1376 */ 1377 static inline bool i_gid_needs_update(struct mnt_idmap *idmap, 1378 const struct iattr *attr, 1379 const struct inode *inode) 1380 { 1381 return ((attr->ia_valid & ATTR_GID) && 1382 !vfsgid_eq(attr->ia_vfsgid, 1383 i_gid_into_vfsgid(idmap, inode))); 1384 } 1385 1386 /** 1387 * i_gid_update - update @inode's i_gid field 1388 * @idmap: idmap of the mount the inode was found from 1389 * @attr: the new attributes of @inode 1390 * @inode: the inode to update 1391 * 1392 * Safely update @inode's i_gid field translating the vfsgid of any idmapped 1393 * mount into the filesystem kgid. 1394 */ 1395 static inline void i_gid_update(struct mnt_idmap *idmap, 1396 const struct iattr *attr, 1397 struct inode *inode) 1398 { 1399 if (attr->ia_valid & ATTR_GID) 1400 inode->i_gid = from_vfsgid(idmap, i_user_ns(inode), 1401 attr->ia_vfsgid); 1402 } 1403 1404 /** 1405 * inode_fsuid_set - initialize inode's i_uid field with callers fsuid 1406 * @inode: inode to initialize 1407 * @idmap: idmap of the mount the inode was found from 1408 * 1409 * Initialize the i_uid field of @inode. If the inode was found/created via 1410 * an idmapped mount map the caller's fsuid according to @idmap. 1411 */ 1412 static inline void inode_fsuid_set(struct inode *inode, 1413 struct mnt_idmap *idmap) 1414 { 1415 inode->i_uid = mapped_fsuid(idmap, i_user_ns(inode)); 1416 } 1417 1418 /** 1419 * inode_fsgid_set - initialize inode's i_gid field with callers fsgid 1420 * @inode: inode to initialize 1421 * @idmap: idmap of the mount the inode was found from 1422 * 1423 * Initialize the i_gid field of @inode. If the inode was found/created via 1424 * an idmapped mount map the caller's fsgid according to @idmap. 1425 */ 1426 static inline void inode_fsgid_set(struct inode *inode, 1427 struct mnt_idmap *idmap) 1428 { 1429 inode->i_gid = mapped_fsgid(idmap, i_user_ns(inode)); 1430 } 1431 1432 /** 1433 * fsuidgid_has_mapping() - check whether caller's fsuid/fsgid is mapped 1434 * @sb: the superblock we want a mapping in 1435 * @idmap: idmap of the relevant mount 1436 * 1437 * Check whether the caller's fsuid and fsgid have a valid mapping in the 1438 * s_user_ns of the superblock @sb. If the caller is on an idmapped mount map 1439 * the caller's fsuid and fsgid according to the @idmap first. 1440 * 1441 * Return: true if fsuid and fsgid is mapped, false if not. 1442 */ 1443 static inline bool fsuidgid_has_mapping(struct super_block *sb, 1444 struct mnt_idmap *idmap) 1445 { 1446 struct user_namespace *fs_userns = sb->s_user_ns; 1447 kuid_t kuid; 1448 kgid_t kgid; 1449 1450 kuid = mapped_fsuid(idmap, fs_userns); 1451 if (!uid_valid(kuid)) 1452 return false; 1453 kgid = mapped_fsgid(idmap, fs_userns); 1454 if (!gid_valid(kgid)) 1455 return false; 1456 return kuid_has_mapping(fs_userns, kuid) && 1457 kgid_has_mapping(fs_userns, kgid); 1458 } 1459 1460 extern struct timespec64 current_time(struct inode *inode); 1461 1462 /* 1463 * Snapshotting support. 1464 */ 1465 1466 /* 1467 * These are internal functions, please use sb_start_{write,pagefault,intwrite} 1468 * instead. 1469 */ 1470 static inline void __sb_end_write(struct super_block *sb, int level) 1471 { 1472 percpu_up_read(sb->s_writers.rw_sem + level-1); 1473 } 1474 1475 static inline void __sb_start_write(struct super_block *sb, int level) 1476 { 1477 percpu_down_read(sb->s_writers.rw_sem + level - 1); 1478 } 1479 1480 static inline bool __sb_start_write_trylock(struct super_block *sb, int level) 1481 { 1482 return percpu_down_read_trylock(sb->s_writers.rw_sem + level - 1); 1483 } 1484 1485 #define __sb_writers_acquired(sb, lev) \ 1486 percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_) 1487 #define __sb_writers_release(sb, lev) \ 1488 percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_) 1489 1490 static inline bool sb_write_started(const struct super_block *sb) 1491 { 1492 return lockdep_is_held_type(sb->s_writers.rw_sem + SB_FREEZE_WRITE - 1, 1); 1493 } 1494 1495 /** 1496 * sb_end_write - drop write access to a superblock 1497 * @sb: the super we wrote to 1498 * 1499 * Decrement number of writers to the filesystem. Wake up possible waiters 1500 * wanting to freeze the filesystem. 1501 */ 1502 static inline void sb_end_write(struct super_block *sb) 1503 { 1504 __sb_end_write(sb, SB_FREEZE_WRITE); 1505 } 1506 1507 /** 1508 * sb_end_pagefault - drop write access to a superblock from a page fault 1509 * @sb: the super we wrote to 1510 * 1511 * Decrement number of processes handling write page fault to the filesystem. 1512 * Wake up possible waiters wanting to freeze the filesystem. 1513 */ 1514 static inline void sb_end_pagefault(struct super_block *sb) 1515 { 1516 __sb_end_write(sb, SB_FREEZE_PAGEFAULT); 1517 } 1518 1519 /** 1520 * sb_end_intwrite - drop write access to a superblock for internal fs purposes 1521 * @sb: the super we wrote to 1522 * 1523 * Decrement fs-internal number of writers to the filesystem. Wake up possible 1524 * waiters wanting to freeze the filesystem. 1525 */ 1526 static inline void sb_end_intwrite(struct super_block *sb) 1527 { 1528 __sb_end_write(sb, SB_FREEZE_FS); 1529 } 1530 1531 /** 1532 * sb_start_write - get write access to a superblock 1533 * @sb: the super we write to 1534 * 1535 * When a process wants to write data or metadata to a file system (i.e. dirty 1536 * a page or an inode), it should embed the operation in a sb_start_write() - 1537 * sb_end_write() pair to get exclusion against file system freezing. This 1538 * function increments number of writers preventing freezing. If the file 1539 * system is already frozen, the function waits until the file system is 1540 * thawed. 1541 * 1542 * Since freeze protection behaves as a lock, users have to preserve 1543 * ordering of freeze protection and other filesystem locks. Generally, 1544 * freeze protection should be the outermost lock. In particular, we have: 1545 * 1546 * sb_start_write 1547 * -> i_mutex (write path, truncate, directory ops, ...) 1548 * -> s_umount (freeze_super, thaw_super) 1549 */ 1550 static inline void sb_start_write(struct super_block *sb) 1551 { 1552 __sb_start_write(sb, SB_FREEZE_WRITE); 1553 } 1554 1555 static inline bool sb_start_write_trylock(struct super_block *sb) 1556 { 1557 return __sb_start_write_trylock(sb, SB_FREEZE_WRITE); 1558 } 1559 1560 /** 1561 * sb_start_pagefault - get write access to a superblock from a page fault 1562 * @sb: the super we write to 1563 * 1564 * When a process starts handling write page fault, it should embed the 1565 * operation into sb_start_pagefault() - sb_end_pagefault() pair to get 1566 * exclusion against file system freezing. This is needed since the page fault 1567 * is going to dirty a page. This function increments number of running page 1568 * faults preventing freezing. If the file system is already frozen, the 1569 * function waits until the file system is thawed. 1570 * 1571 * Since page fault freeze protection behaves as a lock, users have to preserve 1572 * ordering of freeze protection and other filesystem locks. It is advised to 1573 * put sb_start_pagefault() close to mmap_lock in lock ordering. Page fault 1574 * handling code implies lock dependency: 1575 * 1576 * mmap_lock 1577 * -> sb_start_pagefault 1578 */ 1579 static inline void sb_start_pagefault(struct super_block *sb) 1580 { 1581 __sb_start_write(sb, SB_FREEZE_PAGEFAULT); 1582 } 1583 1584 /** 1585 * sb_start_intwrite - get write access to a superblock for internal fs purposes 1586 * @sb: the super we write to 1587 * 1588 * This is the third level of protection against filesystem freezing. It is 1589 * free for use by a filesystem. The only requirement is that it must rank 1590 * below sb_start_pagefault. 1591 * 1592 * For example filesystem can call sb_start_intwrite() when starting a 1593 * transaction which somewhat eases handling of freezing for internal sources 1594 * of filesystem changes (internal fs threads, discarding preallocation on file 1595 * close, etc.). 1596 */ 1597 static inline void sb_start_intwrite(struct super_block *sb) 1598 { 1599 __sb_start_write(sb, SB_FREEZE_FS); 1600 } 1601 1602 static inline bool sb_start_intwrite_trylock(struct super_block *sb) 1603 { 1604 return __sb_start_write_trylock(sb, SB_FREEZE_FS); 1605 } 1606 1607 bool inode_owner_or_capable(struct mnt_idmap *idmap, 1608 const struct inode *inode); 1609 1610 /* 1611 * VFS helper functions.. 1612 */ 1613 int vfs_create(struct mnt_idmap *, struct inode *, 1614 struct dentry *, umode_t, bool); 1615 int vfs_mkdir(struct mnt_idmap *, struct inode *, 1616 struct dentry *, umode_t); 1617 int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *, 1618 umode_t, dev_t); 1619 int vfs_symlink(struct mnt_idmap *, struct inode *, 1620 struct dentry *, const char *); 1621 int vfs_link(struct dentry *, struct mnt_idmap *, struct inode *, 1622 struct dentry *, struct inode **); 1623 int vfs_rmdir(struct mnt_idmap *, struct inode *, struct dentry *); 1624 int vfs_unlink(struct mnt_idmap *, struct inode *, struct dentry *, 1625 struct inode **); 1626 1627 /** 1628 * struct renamedata - contains all information required for renaming 1629 * @old_mnt_idmap: idmap of the old mount the inode was found from 1630 * @old_dir: parent of source 1631 * @old_dentry: source 1632 * @new_mnt_idmap: idmap of the new mount the inode was found from 1633 * @new_dir: parent of destination 1634 * @new_dentry: destination 1635 * @delegated_inode: returns an inode needing a delegation break 1636 * @flags: rename flags 1637 */ 1638 struct renamedata { 1639 struct mnt_idmap *old_mnt_idmap; 1640 struct inode *old_dir; 1641 struct dentry *old_dentry; 1642 struct mnt_idmap *new_mnt_idmap; 1643 struct inode *new_dir; 1644 struct dentry *new_dentry; 1645 struct inode **delegated_inode; 1646 unsigned int flags; 1647 } __randomize_layout; 1648 1649 int vfs_rename(struct renamedata *); 1650 1651 static inline int vfs_whiteout(struct mnt_idmap *idmap, 1652 struct inode *dir, struct dentry *dentry) 1653 { 1654 return vfs_mknod(idmap, dir, dentry, S_IFCHR | WHITEOUT_MODE, 1655 WHITEOUT_DEV); 1656 } 1657 1658 struct file *vfs_tmpfile_open(struct mnt_idmap *idmap, 1659 const struct path *parentpath, 1660 umode_t mode, int open_flag, const struct cred *cred); 1661 1662 int vfs_mkobj(struct dentry *, umode_t, 1663 int (*f)(struct dentry *, umode_t, void *), 1664 void *); 1665 1666 int vfs_fchown(struct file *file, uid_t user, gid_t group); 1667 int vfs_fchmod(struct file *file, umode_t mode); 1668 int vfs_utimes(const struct path *path, struct timespec64 *times); 1669 1670 extern long vfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 1671 1672 #ifdef CONFIG_COMPAT 1673 extern long compat_ptr_ioctl(struct file *file, unsigned int cmd, 1674 unsigned long arg); 1675 #else 1676 #define compat_ptr_ioctl NULL 1677 #endif 1678 1679 /* 1680 * VFS file helper functions. 1681 */ 1682 void inode_init_owner(struct mnt_idmap *idmap, struct inode *inode, 1683 const struct inode *dir, umode_t mode); 1684 extern bool may_open_dev(const struct path *path); 1685 umode_t mode_strip_sgid(struct mnt_idmap *idmap, 1686 const struct inode *dir, umode_t mode); 1687 1688 /* 1689 * This is the "filldir" function type, used by readdir() to let 1690 * the kernel specify what kind of dirent layout it wants to have. 1691 * This allows the kernel to read directories into kernel space or 1692 * to have different dirent layouts depending on the binary type. 1693 * Return 'true' to keep going and 'false' if there are no more entries. 1694 */ 1695 struct dir_context; 1696 typedef bool (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64, 1697 unsigned); 1698 1699 struct dir_context { 1700 filldir_t actor; 1701 loff_t pos; 1702 }; 1703 1704 /* 1705 * These flags let !MMU mmap() govern direct device mapping vs immediate 1706 * copying more easily for MAP_PRIVATE, especially for ROM filesystems. 1707 * 1708 * NOMMU_MAP_COPY: Copy can be mapped (MAP_PRIVATE) 1709 * NOMMU_MAP_DIRECT: Can be mapped directly (MAP_SHARED) 1710 * NOMMU_MAP_READ: Can be mapped for reading 1711 * NOMMU_MAP_WRITE: Can be mapped for writing 1712 * NOMMU_MAP_EXEC: Can be mapped for execution 1713 */ 1714 #define NOMMU_MAP_COPY 0x00000001 1715 #define NOMMU_MAP_DIRECT 0x00000008 1716 #define NOMMU_MAP_READ VM_MAYREAD 1717 #define NOMMU_MAP_WRITE VM_MAYWRITE 1718 #define NOMMU_MAP_EXEC VM_MAYEXEC 1719 1720 #define NOMMU_VMFLAGS \ 1721 (NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC) 1722 1723 /* 1724 * These flags control the behavior of the remap_file_range function pointer. 1725 * If it is called with len == 0 that means "remap to end of source file". 1726 * See Documentation/filesystems/vfs.rst for more details about this call. 1727 * 1728 * REMAP_FILE_DEDUP: only remap if contents identical (i.e. deduplicate) 1729 * REMAP_FILE_CAN_SHORTEN: caller can handle a shortened request 1730 */ 1731 #define REMAP_FILE_DEDUP (1 << 0) 1732 #define REMAP_FILE_CAN_SHORTEN (1 << 1) 1733 1734 /* 1735 * These flags signal that the caller is ok with altering various aspects of 1736 * the behavior of the remap operation. The changes must be made by the 1737 * implementation; the vfs remap helper functions can take advantage of them. 1738 * Flags in this category exist to preserve the quirky behavior of the hoisted 1739 * btrfs clone/dedupe ioctls. 1740 */ 1741 #define REMAP_FILE_ADVISORY (REMAP_FILE_CAN_SHORTEN) 1742 1743 /* 1744 * These flags control the behavior of vfs_copy_file_range(). 1745 * They are not available to the user via syscall. 1746 * 1747 * COPY_FILE_SPLICE: call splice direct instead of fs clone/copy ops 1748 */ 1749 #define COPY_FILE_SPLICE (1 << 0) 1750 1751 struct iov_iter; 1752 struct io_uring_cmd; 1753 1754 struct file_operations { 1755 struct module *owner; 1756 loff_t (*llseek) (struct file *, loff_t, int); 1757 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *); 1758 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *); 1759 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *); 1760 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *); 1761 int (*iopoll)(struct kiocb *kiocb, struct io_comp_batch *, 1762 unsigned int flags); 1763 int (*iterate) (struct file *, struct dir_context *); 1764 int (*iterate_shared) (struct file *, struct dir_context *); 1765 __poll_t (*poll) (struct file *, struct poll_table_struct *); 1766 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long); 1767 long (*compat_ioctl) (struct file *, unsigned int, unsigned long); 1768 int (*mmap) (struct file *, struct vm_area_struct *); 1769 unsigned long mmap_supported_flags; 1770 int (*open) (struct inode *, struct file *); 1771 int (*flush) (struct file *, fl_owner_t id); 1772 int (*release) (struct inode *, struct file *); 1773 int (*fsync) (struct file *, loff_t, loff_t, int datasync); 1774 int (*fasync) (int, struct file *, int); 1775 int (*lock) (struct file *, int, struct file_lock *); 1776 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int); 1777 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 1778 int (*check_flags)(int); 1779 int (*flock) (struct file *, int, struct file_lock *); 1780 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int); 1781 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int); 1782 int (*setlease)(struct file *, long, struct file_lock **, void **); 1783 long (*fallocate)(struct file *file, int mode, loff_t offset, 1784 loff_t len); 1785 void (*show_fdinfo)(struct seq_file *m, struct file *f); 1786 #ifndef CONFIG_MMU 1787 unsigned (*mmap_capabilities)(struct file *); 1788 #endif 1789 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *, 1790 loff_t, size_t, unsigned int); 1791 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in, 1792 struct file *file_out, loff_t pos_out, 1793 loff_t len, unsigned int remap_flags); 1794 int (*fadvise)(struct file *, loff_t, loff_t, int); 1795 int (*uring_cmd)(struct io_uring_cmd *ioucmd, unsigned int issue_flags); 1796 int (*uring_cmd_iopoll)(struct io_uring_cmd *, struct io_comp_batch *, 1797 unsigned int poll_flags); 1798 } __randomize_layout; 1799 1800 struct inode_operations { 1801 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int); 1802 const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *); 1803 int (*permission) (struct mnt_idmap *, struct inode *, int); 1804 struct posix_acl * (*get_inode_acl)(struct inode *, int, bool); 1805 1806 int (*readlink) (struct dentry *, char __user *,int); 1807 1808 int (*create) (struct mnt_idmap *, struct inode *,struct dentry *, 1809 umode_t, bool); 1810 int (*link) (struct dentry *,struct inode *,struct dentry *); 1811 int (*unlink) (struct inode *,struct dentry *); 1812 int (*symlink) (struct mnt_idmap *, struct inode *,struct dentry *, 1813 const char *); 1814 int (*mkdir) (struct mnt_idmap *, struct inode *,struct dentry *, 1815 umode_t); 1816 int (*rmdir) (struct inode *,struct dentry *); 1817 int (*mknod) (struct mnt_idmap *, struct inode *,struct dentry *, 1818 umode_t,dev_t); 1819 int (*rename) (struct mnt_idmap *, struct inode *, struct dentry *, 1820 struct inode *, struct dentry *, unsigned int); 1821 int (*setattr) (struct mnt_idmap *, struct dentry *, struct iattr *); 1822 int (*getattr) (struct mnt_idmap *, const struct path *, 1823 struct kstat *, u32, unsigned int); 1824 ssize_t (*listxattr) (struct dentry *, char *, size_t); 1825 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start, 1826 u64 len); 1827 int (*update_time)(struct inode *, struct timespec64 *, int); 1828 int (*atomic_open)(struct inode *, struct dentry *, 1829 struct file *, unsigned open_flag, 1830 umode_t create_mode); 1831 int (*tmpfile) (struct mnt_idmap *, struct inode *, 1832 struct file *, umode_t); 1833 struct posix_acl *(*get_acl)(struct mnt_idmap *, struct dentry *, 1834 int); 1835 int (*set_acl)(struct mnt_idmap *, struct dentry *, 1836 struct posix_acl *, int); 1837 int (*fileattr_set)(struct mnt_idmap *idmap, 1838 struct dentry *dentry, struct fileattr *fa); 1839 int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa); 1840 } ____cacheline_aligned; 1841 1842 static inline ssize_t call_read_iter(struct file *file, struct kiocb *kio, 1843 struct iov_iter *iter) 1844 { 1845 return file->f_op->read_iter(kio, iter); 1846 } 1847 1848 static inline ssize_t call_write_iter(struct file *file, struct kiocb *kio, 1849 struct iov_iter *iter) 1850 { 1851 return file->f_op->write_iter(kio, iter); 1852 } 1853 1854 static inline int call_mmap(struct file *file, struct vm_area_struct *vma) 1855 { 1856 return file->f_op->mmap(file, vma); 1857 } 1858 1859 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *); 1860 extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *); 1861 extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *, 1862 loff_t, size_t, unsigned int); 1863 extern ssize_t generic_copy_file_range(struct file *file_in, loff_t pos_in, 1864 struct file *file_out, loff_t pos_out, 1865 size_t len, unsigned int flags); 1866 int __generic_remap_file_range_prep(struct file *file_in, loff_t pos_in, 1867 struct file *file_out, loff_t pos_out, 1868 loff_t *len, unsigned int remap_flags, 1869 const struct iomap_ops *dax_read_ops); 1870 int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in, 1871 struct file *file_out, loff_t pos_out, 1872 loff_t *count, unsigned int remap_flags); 1873 extern loff_t do_clone_file_range(struct file *file_in, loff_t pos_in, 1874 struct file *file_out, loff_t pos_out, 1875 loff_t len, unsigned int remap_flags); 1876 extern loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in, 1877 struct file *file_out, loff_t pos_out, 1878 loff_t len, unsigned int remap_flags); 1879 extern int vfs_dedupe_file_range(struct file *file, 1880 struct file_dedupe_range *same); 1881 extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos, 1882 struct file *dst_file, loff_t dst_pos, 1883 loff_t len, unsigned int remap_flags); 1884 1885 1886 struct super_operations { 1887 struct inode *(*alloc_inode)(struct super_block *sb); 1888 void (*destroy_inode)(struct inode *); 1889 void (*free_inode)(struct inode *); 1890 1891 void (*dirty_inode) (struct inode *, int flags); 1892 int (*write_inode) (struct inode *, struct writeback_control *wbc); 1893 int (*drop_inode) (struct inode *); 1894 void (*evict_inode) (struct inode *); 1895 void (*put_super) (struct super_block *); 1896 int (*sync_fs)(struct super_block *sb, int wait); 1897 int (*freeze_super) (struct super_block *); 1898 int (*freeze_fs) (struct super_block *); 1899 int (*thaw_super) (struct super_block *); 1900 int (*unfreeze_fs) (struct super_block *); 1901 int (*statfs) (struct dentry *, struct kstatfs *); 1902 int (*remount_fs) (struct super_block *, int *, char *); 1903 void (*umount_begin) (struct super_block *); 1904 1905 int (*show_options)(struct seq_file *, struct dentry *); 1906 int (*show_devname)(struct seq_file *, struct dentry *); 1907 int (*show_path)(struct seq_file *, struct dentry *); 1908 int (*show_stats)(struct seq_file *, struct dentry *); 1909 #ifdef CONFIG_QUOTA 1910 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t); 1911 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t); 1912 struct dquot **(*get_dquots)(struct inode *); 1913 #endif 1914 long (*nr_cached_objects)(struct super_block *, 1915 struct shrink_control *); 1916 long (*free_cached_objects)(struct super_block *, 1917 struct shrink_control *); 1918 }; 1919 1920 /* 1921 * Inode flags - they have no relation to superblock flags now 1922 */ 1923 #define S_SYNC (1 << 0) /* Writes are synced at once */ 1924 #define S_NOATIME (1 << 1) /* Do not update access times */ 1925 #define S_APPEND (1 << 2) /* Append-only file */ 1926 #define S_IMMUTABLE (1 << 3) /* Immutable file */ 1927 #define S_DEAD (1 << 4) /* removed, but still open directory */ 1928 #define S_NOQUOTA (1 << 5) /* Inode is not counted to quota */ 1929 #define S_DIRSYNC (1 << 6) /* Directory modifications are synchronous */ 1930 #define S_NOCMTIME (1 << 7) /* Do not update file c/mtime */ 1931 #define S_SWAPFILE (1 << 8) /* Do not truncate: swapon got its bmaps */ 1932 #define S_PRIVATE (1 << 9) /* Inode is fs-internal */ 1933 #define S_IMA (1 << 10) /* Inode has an associated IMA struct */ 1934 #define S_AUTOMOUNT (1 << 11) /* Automount/referral quasi-directory */ 1935 #define S_NOSEC (1 << 12) /* no suid or xattr security attributes */ 1936 #ifdef CONFIG_FS_DAX 1937 #define S_DAX (1 << 13) /* Direct Access, avoiding the page cache */ 1938 #else 1939 #define S_DAX 0 /* Make all the DAX code disappear */ 1940 #endif 1941 #define S_ENCRYPTED (1 << 14) /* Encrypted file (using fs/crypto/) */ 1942 #define S_CASEFOLD (1 << 15) /* Casefolded file */ 1943 #define S_VERITY (1 << 16) /* Verity file (using fs/verity/) */ 1944 #define S_KERNEL_FILE (1 << 17) /* File is in use by the kernel (eg. fs/cachefiles) */ 1945 1946 /* 1947 * Note that nosuid etc flags are inode-specific: setting some file-system 1948 * flags just means all the inodes inherit those flags by default. It might be 1949 * possible to override it selectively if you really wanted to with some 1950 * ioctl() that is not currently implemented. 1951 * 1952 * Exception: SB_RDONLY is always applied to the entire file system. 1953 * 1954 * Unfortunately, it is possible to change a filesystems flags with it mounted 1955 * with files in use. This means that all of the inodes will not have their 1956 * i_flags updated. Hence, i_flags no longer inherit the superblock mount 1957 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org 1958 */ 1959 #define __IS_FLG(inode, flg) ((inode)->i_sb->s_flags & (flg)) 1960 1961 static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; } 1962 #define IS_RDONLY(inode) sb_rdonly((inode)->i_sb) 1963 #define IS_SYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS) || \ 1964 ((inode)->i_flags & S_SYNC)) 1965 #define IS_DIRSYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS|SB_DIRSYNC) || \ 1966 ((inode)->i_flags & (S_SYNC|S_DIRSYNC))) 1967 #define IS_MANDLOCK(inode) __IS_FLG(inode, SB_MANDLOCK) 1968 #define IS_NOATIME(inode) __IS_FLG(inode, SB_RDONLY|SB_NOATIME) 1969 #define IS_I_VERSION(inode) __IS_FLG(inode, SB_I_VERSION) 1970 1971 #define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA) 1972 #define IS_APPEND(inode) ((inode)->i_flags & S_APPEND) 1973 #define IS_IMMUTABLE(inode) ((inode)->i_flags & S_IMMUTABLE) 1974 #define IS_POSIXACL(inode) __IS_FLG(inode, SB_POSIXACL) 1975 1976 #define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD) 1977 #define IS_NOCMTIME(inode) ((inode)->i_flags & S_NOCMTIME) 1978 #define IS_SWAPFILE(inode) ((inode)->i_flags & S_SWAPFILE) 1979 #define IS_PRIVATE(inode) ((inode)->i_flags & S_PRIVATE) 1980 #define IS_IMA(inode) ((inode)->i_flags & S_IMA) 1981 #define IS_AUTOMOUNT(inode) ((inode)->i_flags & S_AUTOMOUNT) 1982 #define IS_NOSEC(inode) ((inode)->i_flags & S_NOSEC) 1983 #define IS_DAX(inode) ((inode)->i_flags & S_DAX) 1984 #define IS_ENCRYPTED(inode) ((inode)->i_flags & S_ENCRYPTED) 1985 #define IS_CASEFOLDED(inode) ((inode)->i_flags & S_CASEFOLD) 1986 #define IS_VERITY(inode) ((inode)->i_flags & S_VERITY) 1987 1988 #define IS_WHITEOUT(inode) (S_ISCHR(inode->i_mode) && \ 1989 (inode)->i_rdev == WHITEOUT_DEV) 1990 1991 static inline bool HAS_UNMAPPED_ID(struct mnt_idmap *idmap, 1992 struct inode *inode) 1993 { 1994 return !vfsuid_valid(i_uid_into_vfsuid(idmap, inode)) || 1995 !vfsgid_valid(i_gid_into_vfsgid(idmap, inode)); 1996 } 1997 1998 static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp) 1999 { 2000 *kiocb = (struct kiocb) { 2001 .ki_filp = filp, 2002 .ki_flags = filp->f_iocb_flags, 2003 .ki_ioprio = get_current_ioprio(), 2004 }; 2005 } 2006 2007 static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src, 2008 struct file *filp) 2009 { 2010 *kiocb = (struct kiocb) { 2011 .ki_filp = filp, 2012 .ki_flags = kiocb_src->ki_flags, 2013 .ki_ioprio = kiocb_src->ki_ioprio, 2014 .ki_pos = kiocb_src->ki_pos, 2015 }; 2016 } 2017 2018 /* 2019 * Inode state bits. Protected by inode->i_lock 2020 * 2021 * Four bits determine the dirty state of the inode: I_DIRTY_SYNC, 2022 * I_DIRTY_DATASYNC, I_DIRTY_PAGES, and I_DIRTY_TIME. 2023 * 2024 * Four bits define the lifetime of an inode. Initially, inodes are I_NEW, 2025 * until that flag is cleared. I_WILL_FREE, I_FREEING and I_CLEAR are set at 2026 * various stages of removing an inode. 2027 * 2028 * Two bits are used for locking and completion notification, I_NEW and I_SYNC. 2029 * 2030 * I_DIRTY_SYNC Inode is dirty, but doesn't have to be written on 2031 * fdatasync() (unless I_DIRTY_DATASYNC is also set). 2032 * Timestamp updates are the usual cause. 2033 * I_DIRTY_DATASYNC Data-related inode changes pending. We keep track of 2034 * these changes separately from I_DIRTY_SYNC so that we 2035 * don't have to write inode on fdatasync() when only 2036 * e.g. the timestamps have changed. 2037 * I_DIRTY_PAGES Inode has dirty pages. Inode itself may be clean. 2038 * I_DIRTY_TIME The inode itself has dirty timestamps, and the 2039 * lazytime mount option is enabled. We keep track of this 2040 * separately from I_DIRTY_SYNC in order to implement 2041 * lazytime. This gets cleared if I_DIRTY_INODE 2042 * (I_DIRTY_SYNC and/or I_DIRTY_DATASYNC) gets set. But 2043 * I_DIRTY_TIME can still be set if I_DIRTY_SYNC is already 2044 * in place because writeback might already be in progress 2045 * and we don't want to lose the time update 2046 * I_NEW Serves as both a mutex and completion notification. 2047 * New inodes set I_NEW. If two processes both create 2048 * the same inode, one of them will release its inode and 2049 * wait for I_NEW to be released before returning. 2050 * Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can 2051 * also cause waiting on I_NEW, without I_NEW actually 2052 * being set. find_inode() uses this to prevent returning 2053 * nearly-dead inodes. 2054 * I_WILL_FREE Must be set when calling write_inode_now() if i_count 2055 * is zero. I_FREEING must be set when I_WILL_FREE is 2056 * cleared. 2057 * I_FREEING Set when inode is about to be freed but still has dirty 2058 * pages or buffers attached or the inode itself is still 2059 * dirty. 2060 * I_CLEAR Added by clear_inode(). In this state the inode is 2061 * clean and can be destroyed. Inode keeps I_FREEING. 2062 * 2063 * Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are 2064 * prohibited for many purposes. iget() must wait for 2065 * the inode to be completely released, then create it 2066 * anew. Other functions will just ignore such inodes, 2067 * if appropriate. I_NEW is used for waiting. 2068 * 2069 * I_SYNC Writeback of inode is running. The bit is set during 2070 * data writeback, and cleared with a wakeup on the bit 2071 * address once it is done. The bit is also used to pin 2072 * the inode in memory for flusher thread. 2073 * 2074 * I_REFERENCED Marks the inode as recently references on the LRU list. 2075 * 2076 * I_DIO_WAKEUP Never set. Only used as a key for wait_on_bit(). 2077 * 2078 * I_WB_SWITCH Cgroup bdi_writeback switching in progress. Used to 2079 * synchronize competing switching instances and to tell 2080 * wb stat updates to grab the i_pages lock. See 2081 * inode_switch_wbs_work_fn() for details. 2082 * 2083 * I_OVL_INUSE Used by overlayfs to get exclusive ownership on upper 2084 * and work dirs among overlayfs mounts. 2085 * 2086 * I_CREATING New object's inode in the middle of setting up. 2087 * 2088 * I_DONTCACHE Evict inode as soon as it is not used anymore. 2089 * 2090 * I_SYNC_QUEUED Inode is queued in b_io or b_more_io writeback lists. 2091 * Used to detect that mark_inode_dirty() should not move 2092 * inode between dirty lists. 2093 * 2094 * I_PINNING_FSCACHE_WB Inode is pinning an fscache object for writeback. 2095 * 2096 * Q: What is the difference between I_WILL_FREE and I_FREEING? 2097 */ 2098 #define I_DIRTY_SYNC (1 << 0) 2099 #define I_DIRTY_DATASYNC (1 << 1) 2100 #define I_DIRTY_PAGES (1 << 2) 2101 #define __I_NEW 3 2102 #define I_NEW (1 << __I_NEW) 2103 #define I_WILL_FREE (1 << 4) 2104 #define I_FREEING (1 << 5) 2105 #define I_CLEAR (1 << 6) 2106 #define __I_SYNC 7 2107 #define I_SYNC (1 << __I_SYNC) 2108 #define I_REFERENCED (1 << 8) 2109 #define __I_DIO_WAKEUP 9 2110 #define I_DIO_WAKEUP (1 << __I_DIO_WAKEUP) 2111 #define I_LINKABLE (1 << 10) 2112 #define I_DIRTY_TIME (1 << 11) 2113 #define I_WB_SWITCH (1 << 13) 2114 #define I_OVL_INUSE (1 << 14) 2115 #define I_CREATING (1 << 15) 2116 #define I_DONTCACHE (1 << 16) 2117 #define I_SYNC_QUEUED (1 << 17) 2118 #define I_PINNING_FSCACHE_WB (1 << 18) 2119 2120 #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC) 2121 #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES) 2122 #define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME) 2123 2124 extern void __mark_inode_dirty(struct inode *, int); 2125 static inline void mark_inode_dirty(struct inode *inode) 2126 { 2127 __mark_inode_dirty(inode, I_DIRTY); 2128 } 2129 2130 static inline void mark_inode_dirty_sync(struct inode *inode) 2131 { 2132 __mark_inode_dirty(inode, I_DIRTY_SYNC); 2133 } 2134 2135 /* 2136 * Returns true if the given inode itself only has dirty timestamps (its pages 2137 * may still be dirty) and isn't currently being allocated or freed. 2138 * Filesystems should call this if when writing an inode when lazytime is 2139 * enabled, they want to opportunistically write the timestamps of other inodes 2140 * located very nearby on-disk, e.g. in the same inode block. This returns true 2141 * if the given inode is in need of such an opportunistic update. Requires 2142 * i_lock, or at least later re-checking under i_lock. 2143 */ 2144 static inline bool inode_is_dirtytime_only(struct inode *inode) 2145 { 2146 return (inode->i_state & (I_DIRTY_TIME | I_NEW | 2147 I_FREEING | I_WILL_FREE)) == I_DIRTY_TIME; 2148 } 2149 2150 extern void inc_nlink(struct inode *inode); 2151 extern void drop_nlink(struct inode *inode); 2152 extern void clear_nlink(struct inode *inode); 2153 extern void set_nlink(struct inode *inode, unsigned int nlink); 2154 2155 static inline void inode_inc_link_count(struct inode *inode) 2156 { 2157 inc_nlink(inode); 2158 mark_inode_dirty(inode); 2159 } 2160 2161 static inline void inode_dec_link_count(struct inode *inode) 2162 { 2163 drop_nlink(inode); 2164 mark_inode_dirty(inode); 2165 } 2166 2167 enum file_time_flags { 2168 S_ATIME = 1, 2169 S_MTIME = 2, 2170 S_CTIME = 4, 2171 S_VERSION = 8, 2172 }; 2173 2174 extern bool atime_needs_update(const struct path *, struct inode *); 2175 extern void touch_atime(const struct path *); 2176 int inode_update_time(struct inode *inode, struct timespec64 *time, int flags); 2177 2178 static inline void file_accessed(struct file *file) 2179 { 2180 if (!(file->f_flags & O_NOATIME)) 2181 touch_atime(&file->f_path); 2182 } 2183 2184 extern int file_modified(struct file *file); 2185 int kiocb_modified(struct kiocb *iocb); 2186 2187 int sync_inode_metadata(struct inode *inode, int wait); 2188 2189 struct file_system_type { 2190 const char *name; 2191 int fs_flags; 2192 #define FS_REQUIRES_DEV 1 2193 #define FS_BINARY_MOUNTDATA 2 2194 #define FS_HAS_SUBTYPE 4 2195 #define FS_USERNS_MOUNT 8 /* Can be mounted by userns root */ 2196 #define FS_DISALLOW_NOTIFY_PERM 16 /* Disable fanotify permission events */ 2197 #define FS_ALLOW_IDMAP 32 /* FS has been updated to handle vfs idmappings. */ 2198 #define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() during rename() internally. */ 2199 int (*init_fs_context)(struct fs_context *); 2200 const struct fs_parameter_spec *parameters; 2201 struct dentry *(*mount) (struct file_system_type *, int, 2202 const char *, void *); 2203 void (*kill_sb) (struct super_block *); 2204 struct module *owner; 2205 struct file_system_type * next; 2206 struct hlist_head fs_supers; 2207 2208 struct lock_class_key s_lock_key; 2209 struct lock_class_key s_umount_key; 2210 struct lock_class_key s_vfs_rename_key; 2211 struct lock_class_key s_writers_key[SB_FREEZE_LEVELS]; 2212 2213 struct lock_class_key i_lock_key; 2214 struct lock_class_key i_mutex_key; 2215 struct lock_class_key invalidate_lock_key; 2216 struct lock_class_key i_mutex_dir_key; 2217 }; 2218 2219 #define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME) 2220 2221 extern struct dentry *mount_bdev(struct file_system_type *fs_type, 2222 int flags, const char *dev_name, void *data, 2223 int (*fill_super)(struct super_block *, void *, int)); 2224 extern struct dentry *mount_single(struct file_system_type *fs_type, 2225 int flags, void *data, 2226 int (*fill_super)(struct super_block *, void *, int)); 2227 extern struct dentry *mount_nodev(struct file_system_type *fs_type, 2228 int flags, void *data, 2229 int (*fill_super)(struct super_block *, void *, int)); 2230 extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path); 2231 void retire_super(struct super_block *sb); 2232 void generic_shutdown_super(struct super_block *sb); 2233 void kill_block_super(struct super_block *sb); 2234 void kill_anon_super(struct super_block *sb); 2235 void kill_litter_super(struct super_block *sb); 2236 void deactivate_super(struct super_block *sb); 2237 void deactivate_locked_super(struct super_block *sb); 2238 int set_anon_super(struct super_block *s, void *data); 2239 int set_anon_super_fc(struct super_block *s, struct fs_context *fc); 2240 int get_anon_bdev(dev_t *); 2241 void free_anon_bdev(dev_t); 2242 struct super_block *sget_fc(struct fs_context *fc, 2243 int (*test)(struct super_block *, struct fs_context *), 2244 int (*set)(struct super_block *, struct fs_context *)); 2245 struct super_block *sget(struct file_system_type *type, 2246 int (*test)(struct super_block *,void *), 2247 int (*set)(struct super_block *,void *), 2248 int flags, void *data); 2249 2250 /* Alas, no aliases. Too much hassle with bringing module.h everywhere */ 2251 #define fops_get(fops) \ 2252 (((fops) && try_module_get((fops)->owner) ? (fops) : NULL)) 2253 #define fops_put(fops) \ 2254 do { if (fops) module_put((fops)->owner); } while(0) 2255 /* 2256 * This one is to be used *ONLY* from ->open() instances. 2257 * fops must be non-NULL, pinned down *and* module dependencies 2258 * should be sufficient to pin the caller down as well. 2259 */ 2260 #define replace_fops(f, fops) \ 2261 do { \ 2262 struct file *__file = (f); \ 2263 fops_put(__file->f_op); \ 2264 BUG_ON(!(__file->f_op = (fops))); \ 2265 } while(0) 2266 2267 extern int register_filesystem(struct file_system_type *); 2268 extern int unregister_filesystem(struct file_system_type *); 2269 extern int vfs_statfs(const struct path *, struct kstatfs *); 2270 extern int user_statfs(const char __user *, struct kstatfs *); 2271 extern int fd_statfs(int, struct kstatfs *); 2272 extern int freeze_super(struct super_block *super); 2273 extern int thaw_super(struct super_block *super); 2274 extern __printf(2, 3) 2275 int super_setup_bdi_name(struct super_block *sb, char *fmt, ...); 2276 extern int super_setup_bdi(struct super_block *sb); 2277 2278 extern int current_umask(void); 2279 2280 extern void ihold(struct inode * inode); 2281 extern void iput(struct inode *); 2282 extern int generic_update_time(struct inode *, struct timespec64 *, int); 2283 2284 /* /sys/fs */ 2285 extern struct kobject *fs_kobj; 2286 2287 #define MAX_RW_COUNT (INT_MAX & PAGE_MASK) 2288 2289 /* fs/open.c */ 2290 struct audit_names; 2291 struct filename { 2292 const char *name; /* pointer to actual string */ 2293 const __user char *uptr; /* original userland pointer */ 2294 int refcnt; 2295 struct audit_names *aname; 2296 const char iname[]; 2297 }; 2298 static_assert(offsetof(struct filename, iname) % sizeof(long) == 0); 2299 2300 static inline struct mnt_idmap *file_mnt_idmap(struct file *file) 2301 { 2302 return mnt_idmap(file->f_path.mnt); 2303 } 2304 2305 /** 2306 * is_idmapped_mnt - check whether a mount is mapped 2307 * @mnt: the mount to check 2308 * 2309 * If @mnt has an non @nop_mnt_idmap attached to it then @mnt is mapped. 2310 * 2311 * Return: true if mount is mapped, false if not. 2312 */ 2313 static inline bool is_idmapped_mnt(const struct vfsmount *mnt) 2314 { 2315 return mnt_idmap(mnt) != &nop_mnt_idmap; 2316 } 2317 2318 extern long vfs_truncate(const struct path *, loff_t); 2319 int do_truncate(struct mnt_idmap *, struct dentry *, loff_t start, 2320 unsigned int time_attrs, struct file *filp); 2321 extern int vfs_fallocate(struct file *file, int mode, loff_t offset, 2322 loff_t len); 2323 extern long do_sys_open(int dfd, const char __user *filename, int flags, 2324 umode_t mode); 2325 extern struct file *file_open_name(struct filename *, int, umode_t); 2326 extern struct file *filp_open(const char *, int, umode_t); 2327 extern struct file *file_open_root(const struct path *, 2328 const char *, int, umode_t); 2329 static inline struct file *file_open_root_mnt(struct vfsmount *mnt, 2330 const char *name, int flags, umode_t mode) 2331 { 2332 return file_open_root(&(struct path){.mnt = mnt, .dentry = mnt->mnt_root}, 2333 name, flags, mode); 2334 } 2335 extern struct file * dentry_open(const struct path *, int, const struct cred *); 2336 extern struct file *dentry_create(const struct path *path, int flags, 2337 umode_t mode, const struct cred *cred); 2338 extern struct file * open_with_fake_path(const struct path *, int, 2339 struct inode*, const struct cred *); 2340 static inline struct file *file_clone_open(struct file *file) 2341 { 2342 return dentry_open(&file->f_path, file->f_flags, file->f_cred); 2343 } 2344 extern int filp_close(struct file *, fl_owner_t id); 2345 2346 extern struct filename *getname_flags(const char __user *, int, int *); 2347 extern struct filename *getname_uflags(const char __user *, int); 2348 extern struct filename *getname(const char __user *); 2349 extern struct filename *getname_kernel(const char *); 2350 extern void putname(struct filename *name); 2351 2352 extern int finish_open(struct file *file, struct dentry *dentry, 2353 int (*open)(struct inode *, struct file *)); 2354 extern int finish_no_open(struct file *file, struct dentry *dentry); 2355 2356 /* Helper for the simple case when original dentry is used */ 2357 static inline int finish_open_simple(struct file *file, int error) 2358 { 2359 if (error) 2360 return error; 2361 2362 return finish_open(file, file->f_path.dentry, NULL); 2363 } 2364 2365 /* fs/dcache.c */ 2366 extern void __init vfs_caches_init_early(void); 2367 extern void __init vfs_caches_init(void); 2368 2369 extern struct kmem_cache *names_cachep; 2370 2371 #define __getname() kmem_cache_alloc(names_cachep, GFP_KERNEL) 2372 #define __putname(name) kmem_cache_free(names_cachep, (void *)(name)) 2373 2374 extern struct super_block *blockdev_superblock; 2375 static inline bool sb_is_blkdev_sb(struct super_block *sb) 2376 { 2377 return IS_ENABLED(CONFIG_BLOCK) && sb == blockdev_superblock; 2378 } 2379 2380 void emergency_thaw_all(void); 2381 extern int sync_filesystem(struct super_block *); 2382 extern const struct file_operations def_blk_fops; 2383 extern const struct file_operations def_chr_fops; 2384 2385 /* fs/char_dev.c */ 2386 #define CHRDEV_MAJOR_MAX 512 2387 /* Marks the bottom of the first segment of free char majors */ 2388 #define CHRDEV_MAJOR_DYN_END 234 2389 /* Marks the top and bottom of the second segment of free char majors */ 2390 #define CHRDEV_MAJOR_DYN_EXT_START 511 2391 #define CHRDEV_MAJOR_DYN_EXT_END 384 2392 2393 extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *); 2394 extern int register_chrdev_region(dev_t, unsigned, const char *); 2395 extern int __register_chrdev(unsigned int major, unsigned int baseminor, 2396 unsigned int count, const char *name, 2397 const struct file_operations *fops); 2398 extern void __unregister_chrdev(unsigned int major, unsigned int baseminor, 2399 unsigned int count, const char *name); 2400 extern void unregister_chrdev_region(dev_t, unsigned); 2401 extern void chrdev_show(struct seq_file *,off_t); 2402 2403 static inline int register_chrdev(unsigned int major, const char *name, 2404 const struct file_operations *fops) 2405 { 2406 return __register_chrdev(major, 0, 256, name, fops); 2407 } 2408 2409 static inline void unregister_chrdev(unsigned int major, const char *name) 2410 { 2411 __unregister_chrdev(major, 0, 256, name); 2412 } 2413 2414 extern void init_special_inode(struct inode *, umode_t, dev_t); 2415 2416 /* Invalid inode operations -- fs/bad_inode.c */ 2417 extern void make_bad_inode(struct inode *); 2418 extern bool is_bad_inode(struct inode *); 2419 2420 extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart, 2421 loff_t lend); 2422 extern int __must_check file_check_and_advance_wb_err(struct file *file); 2423 extern int __must_check file_write_and_wait_range(struct file *file, 2424 loff_t start, loff_t end); 2425 2426 static inline int file_write_and_wait(struct file *file) 2427 { 2428 return file_write_and_wait_range(file, 0, LLONG_MAX); 2429 } 2430 2431 extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end, 2432 int datasync); 2433 extern int vfs_fsync(struct file *file, int datasync); 2434 2435 extern int sync_file_range(struct file *file, loff_t offset, loff_t nbytes, 2436 unsigned int flags); 2437 2438 static inline bool iocb_is_dsync(const struct kiocb *iocb) 2439 { 2440 return (iocb->ki_flags & IOCB_DSYNC) || 2441 IS_SYNC(iocb->ki_filp->f_mapping->host); 2442 } 2443 2444 /* 2445 * Sync the bytes written if this was a synchronous write. Expect ki_pos 2446 * to already be updated for the write, and will return either the amount 2447 * of bytes passed in, or an error if syncing the file failed. 2448 */ 2449 static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count) 2450 { 2451 if (iocb_is_dsync(iocb)) { 2452 int ret = vfs_fsync_range(iocb->ki_filp, 2453 iocb->ki_pos - count, iocb->ki_pos - 1, 2454 (iocb->ki_flags & IOCB_SYNC) ? 0 : 1); 2455 if (ret) 2456 return ret; 2457 } 2458 2459 return count; 2460 } 2461 2462 extern void emergency_sync(void); 2463 extern void emergency_remount(void); 2464 2465 #ifdef CONFIG_BLOCK 2466 extern int bmap(struct inode *inode, sector_t *block); 2467 #else 2468 static inline int bmap(struct inode *inode, sector_t *block) 2469 { 2470 return -EINVAL; 2471 } 2472 #endif 2473 2474 int notify_change(struct mnt_idmap *, struct dentry *, 2475 struct iattr *, struct inode **); 2476 int inode_permission(struct mnt_idmap *, struct inode *, int); 2477 int generic_permission(struct mnt_idmap *, struct inode *, int); 2478 static inline int file_permission(struct file *file, int mask) 2479 { 2480 return inode_permission(file_mnt_idmap(file), 2481 file_inode(file), mask); 2482 } 2483 static inline int path_permission(const struct path *path, int mask) 2484 { 2485 return inode_permission(mnt_idmap(path->mnt), 2486 d_inode(path->dentry), mask); 2487 } 2488 int __check_sticky(struct mnt_idmap *idmap, struct inode *dir, 2489 struct inode *inode); 2490 2491 static inline bool execute_ok(struct inode *inode) 2492 { 2493 return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode); 2494 } 2495 2496 static inline bool inode_wrong_type(const struct inode *inode, umode_t mode) 2497 { 2498 return (inode->i_mode ^ mode) & S_IFMT; 2499 } 2500 2501 static inline void file_start_write(struct file *file) 2502 { 2503 if (!S_ISREG(file_inode(file)->i_mode)) 2504 return; 2505 sb_start_write(file_inode(file)->i_sb); 2506 } 2507 2508 static inline bool file_start_write_trylock(struct file *file) 2509 { 2510 if (!S_ISREG(file_inode(file)->i_mode)) 2511 return true; 2512 return sb_start_write_trylock(file_inode(file)->i_sb); 2513 } 2514 2515 static inline void file_end_write(struct file *file) 2516 { 2517 if (!S_ISREG(file_inode(file)->i_mode)) 2518 return; 2519 __sb_end_write(file_inode(file)->i_sb, SB_FREEZE_WRITE); 2520 } 2521 2522 /* 2523 * This is used for regular files where some users -- especially the 2524 * currently executed binary in a process, previously handled via 2525 * VM_DENYWRITE -- cannot handle concurrent write (and maybe mmap 2526 * read-write shared) accesses. 2527 * 2528 * get_write_access() gets write permission for a file. 2529 * put_write_access() releases this write permission. 2530 * deny_write_access() denies write access to a file. 2531 * allow_write_access() re-enables write access to a file. 2532 * 2533 * The i_writecount field of an inode can have the following values: 2534 * 0: no write access, no denied write access 2535 * < 0: (-i_writecount) users that denied write access to the file. 2536 * > 0: (i_writecount) users that have write access to the file. 2537 * 2538 * Normally we operate on that counter with atomic_{inc,dec} and it's safe 2539 * except for the cases where we don't hold i_writecount yet. Then we need to 2540 * use {get,deny}_write_access() - these functions check the sign and refuse 2541 * to do the change if sign is wrong. 2542 */ 2543 static inline int get_write_access(struct inode *inode) 2544 { 2545 return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY; 2546 } 2547 static inline int deny_write_access(struct file *file) 2548 { 2549 struct inode *inode = file_inode(file); 2550 return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY; 2551 } 2552 static inline void put_write_access(struct inode * inode) 2553 { 2554 atomic_dec(&inode->i_writecount); 2555 } 2556 static inline void allow_write_access(struct file *file) 2557 { 2558 if (file) 2559 atomic_inc(&file_inode(file)->i_writecount); 2560 } 2561 static inline bool inode_is_open_for_write(const struct inode *inode) 2562 { 2563 return atomic_read(&inode->i_writecount) > 0; 2564 } 2565 2566 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING) 2567 static inline void i_readcount_dec(struct inode *inode) 2568 { 2569 BUG_ON(!atomic_read(&inode->i_readcount)); 2570 atomic_dec(&inode->i_readcount); 2571 } 2572 static inline void i_readcount_inc(struct inode *inode) 2573 { 2574 atomic_inc(&inode->i_readcount); 2575 } 2576 #else 2577 static inline void i_readcount_dec(struct inode *inode) 2578 { 2579 return; 2580 } 2581 static inline void i_readcount_inc(struct inode *inode) 2582 { 2583 return; 2584 } 2585 #endif 2586 extern int do_pipe_flags(int *, int); 2587 2588 extern ssize_t kernel_read(struct file *, void *, size_t, loff_t *); 2589 ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos); 2590 extern ssize_t kernel_write(struct file *, const void *, size_t, loff_t *); 2591 extern ssize_t __kernel_write(struct file *, const void *, size_t, loff_t *); 2592 extern struct file * open_exec(const char *); 2593 2594 /* fs/dcache.c -- generic fs support functions */ 2595 extern bool is_subdir(struct dentry *, struct dentry *); 2596 extern bool path_is_under(const struct path *, const struct path *); 2597 2598 extern char *file_path(struct file *, char *, int); 2599 2600 #include <linux/err.h> 2601 2602 /* needed for stackable file system support */ 2603 extern loff_t default_llseek(struct file *file, loff_t offset, int whence); 2604 2605 extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence); 2606 2607 extern int inode_init_always(struct super_block *, struct inode *); 2608 extern void inode_init_once(struct inode *); 2609 extern void address_space_init_once(struct address_space *mapping); 2610 extern struct inode * igrab(struct inode *); 2611 extern ino_t iunique(struct super_block *, ino_t); 2612 extern int inode_needs_sync(struct inode *inode); 2613 extern int generic_delete_inode(struct inode *inode); 2614 static inline int generic_drop_inode(struct inode *inode) 2615 { 2616 return !inode->i_nlink || inode_unhashed(inode); 2617 } 2618 extern void d_mark_dontcache(struct inode *inode); 2619 2620 extern struct inode *ilookup5_nowait(struct super_block *sb, 2621 unsigned long hashval, int (*test)(struct inode *, void *), 2622 void *data); 2623 extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval, 2624 int (*test)(struct inode *, void *), void *data); 2625 extern struct inode *ilookup(struct super_block *sb, unsigned long ino); 2626 2627 extern struct inode *inode_insert5(struct inode *inode, unsigned long hashval, 2628 int (*test)(struct inode *, void *), 2629 int (*set)(struct inode *, void *), 2630 void *data); 2631 extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *); 2632 extern struct inode * iget_locked(struct super_block *, unsigned long); 2633 extern struct inode *find_inode_nowait(struct super_block *, 2634 unsigned long, 2635 int (*match)(struct inode *, 2636 unsigned long, void *), 2637 void *data); 2638 extern struct inode *find_inode_rcu(struct super_block *, unsigned long, 2639 int (*)(struct inode *, void *), void *); 2640 extern struct inode *find_inode_by_ino_rcu(struct super_block *, unsigned long); 2641 extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *); 2642 extern int insert_inode_locked(struct inode *); 2643 #ifdef CONFIG_DEBUG_LOCK_ALLOC 2644 extern void lockdep_annotate_inode_mutex_key(struct inode *inode); 2645 #else 2646 static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { }; 2647 #endif 2648 extern void unlock_new_inode(struct inode *); 2649 extern void discard_new_inode(struct inode *); 2650 extern unsigned int get_next_ino(void); 2651 extern void evict_inodes(struct super_block *sb); 2652 void dump_mapping(const struct address_space *); 2653 2654 /* 2655 * Userspace may rely on the the inode number being non-zero. For example, glibc 2656 * simply ignores files with zero i_ino in unlink() and other places. 2657 * 2658 * As an additional complication, if userspace was compiled with 2659 * _FILE_OFFSET_BITS=32 on a 64-bit kernel we'll only end up reading out the 2660 * lower 32 bits, so we need to check that those aren't zero explicitly. With 2661 * _FILE_OFFSET_BITS=64, this may cause some harmless false-negatives, but 2662 * better safe than sorry. 2663 */ 2664 static inline bool is_zero_ino(ino_t ino) 2665 { 2666 return (u32)ino == 0; 2667 } 2668 2669 extern void __iget(struct inode * inode); 2670 extern void iget_failed(struct inode *); 2671 extern void clear_inode(struct inode *); 2672 extern void __destroy_inode(struct inode *); 2673 extern struct inode *new_inode_pseudo(struct super_block *sb); 2674 extern struct inode *new_inode(struct super_block *sb); 2675 extern void free_inode_nonrcu(struct inode *inode); 2676 extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *); 2677 extern int file_remove_privs(struct file *); 2678 2679 /* 2680 * This must be used for allocating filesystems specific inodes to set 2681 * up the inode reclaim context correctly. 2682 */ 2683 static inline void * 2684 alloc_inode_sb(struct super_block *sb, struct kmem_cache *cache, gfp_t gfp) 2685 { 2686 return kmem_cache_alloc_lru(cache, &sb->s_inode_lru, gfp); 2687 } 2688 2689 extern void __insert_inode_hash(struct inode *, unsigned long hashval); 2690 static inline void insert_inode_hash(struct inode *inode) 2691 { 2692 __insert_inode_hash(inode, inode->i_ino); 2693 } 2694 2695 extern void __remove_inode_hash(struct inode *); 2696 static inline void remove_inode_hash(struct inode *inode) 2697 { 2698 if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash)) 2699 __remove_inode_hash(inode); 2700 } 2701 2702 extern void inode_sb_list_add(struct inode *inode); 2703 extern void inode_add_lru(struct inode *inode); 2704 2705 extern int sb_set_blocksize(struct super_block *, int); 2706 extern int sb_min_blocksize(struct super_block *, int); 2707 2708 extern int generic_file_mmap(struct file *, struct vm_area_struct *); 2709 extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *); 2710 extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *); 2711 int generic_write_checks_count(struct kiocb *iocb, loff_t *count); 2712 extern int generic_write_check_limits(struct file *file, loff_t pos, 2713 loff_t *count); 2714 extern int generic_file_rw_checks(struct file *file_in, struct file *file_out); 2715 ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *to, 2716 ssize_t already_read); 2717 extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *); 2718 extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *); 2719 extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *); 2720 extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *); 2721 ssize_t generic_perform_write(struct kiocb *, struct iov_iter *); 2722 2723 ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos, 2724 rwf_t flags); 2725 ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos, 2726 rwf_t flags); 2727 ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb, 2728 struct iov_iter *iter); 2729 ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb, 2730 struct iov_iter *iter); 2731 2732 /* fs/splice.c */ 2733 ssize_t filemap_splice_read(struct file *in, loff_t *ppos, 2734 struct pipe_inode_info *pipe, 2735 size_t len, unsigned int flags); 2736 ssize_t direct_splice_read(struct file *in, loff_t *ppos, 2737 struct pipe_inode_info *pipe, 2738 size_t len, unsigned int flags); 2739 extern ssize_t generic_file_splice_read(struct file *, loff_t *, 2740 struct pipe_inode_info *, size_t, unsigned int); 2741 extern ssize_t iter_file_splice_write(struct pipe_inode_info *, 2742 struct file *, loff_t *, size_t, unsigned int); 2743 extern ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, 2744 struct file *out, loff_t *, size_t len, unsigned int flags); 2745 extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out, 2746 loff_t *opos, size_t len, unsigned int flags); 2747 2748 2749 extern void 2750 file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping); 2751 extern loff_t noop_llseek(struct file *file, loff_t offset, int whence); 2752 #define no_llseek NULL 2753 extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize); 2754 extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence); 2755 extern loff_t generic_file_llseek_size(struct file *file, loff_t offset, 2756 int whence, loff_t maxsize, loff_t eof); 2757 extern loff_t fixed_size_llseek(struct file *file, loff_t offset, 2758 int whence, loff_t size); 2759 extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t); 2760 extern loff_t no_seek_end_llseek(struct file *, loff_t, int); 2761 int rw_verify_area(int, struct file *, const loff_t *, size_t); 2762 extern int generic_file_open(struct inode * inode, struct file * filp); 2763 extern int nonseekable_open(struct inode * inode, struct file * filp); 2764 extern int stream_open(struct inode * inode, struct file * filp); 2765 2766 #ifdef CONFIG_BLOCK 2767 typedef void (dio_submit_t)(struct bio *bio, struct inode *inode, 2768 loff_t file_offset); 2769 2770 enum { 2771 /* need locking between buffered and direct access */ 2772 DIO_LOCKING = 0x01, 2773 2774 /* filesystem does not support filling holes */ 2775 DIO_SKIP_HOLES = 0x02, 2776 }; 2777 2778 ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode, 2779 struct block_device *bdev, struct iov_iter *iter, 2780 get_block_t get_block, 2781 dio_iodone_t end_io, dio_submit_t submit_io, 2782 int flags); 2783 2784 static inline ssize_t blockdev_direct_IO(struct kiocb *iocb, 2785 struct inode *inode, 2786 struct iov_iter *iter, 2787 get_block_t get_block) 2788 { 2789 return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter, 2790 get_block, NULL, NULL, DIO_LOCKING | DIO_SKIP_HOLES); 2791 } 2792 #endif 2793 2794 void inode_dio_wait(struct inode *inode); 2795 2796 /** 2797 * inode_dio_begin - signal start of a direct I/O requests 2798 * @inode: inode the direct I/O happens on 2799 * 2800 * This is called once we've finished processing a direct I/O request, 2801 * and is used to wake up callers waiting for direct I/O to be quiesced. 2802 */ 2803 static inline void inode_dio_begin(struct inode *inode) 2804 { 2805 atomic_inc(&inode->i_dio_count); 2806 } 2807 2808 /** 2809 * inode_dio_end - signal finish of a direct I/O requests 2810 * @inode: inode the direct I/O happens on 2811 * 2812 * This is called once we've finished processing a direct I/O request, 2813 * and is used to wake up callers waiting for direct I/O to be quiesced. 2814 */ 2815 static inline void inode_dio_end(struct inode *inode) 2816 { 2817 if (atomic_dec_and_test(&inode->i_dio_count)) 2818 wake_up_bit(&inode->i_state, __I_DIO_WAKEUP); 2819 } 2820 2821 /* 2822 * Warn about a page cache invalidation failure diring a direct I/O write. 2823 */ 2824 void dio_warn_stale_pagecache(struct file *filp); 2825 2826 extern void inode_set_flags(struct inode *inode, unsigned int flags, 2827 unsigned int mask); 2828 2829 extern const struct file_operations generic_ro_fops; 2830 2831 #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m)) 2832 2833 extern int readlink_copy(char __user *, int, const char *); 2834 extern int page_readlink(struct dentry *, char __user *, int); 2835 extern const char *page_get_link(struct dentry *, struct inode *, 2836 struct delayed_call *); 2837 extern void page_put_link(void *); 2838 extern int page_symlink(struct inode *inode, const char *symname, int len); 2839 extern const struct inode_operations page_symlink_inode_operations; 2840 extern void kfree_link(void *); 2841 void generic_fillattr(struct mnt_idmap *, struct inode *, struct kstat *); 2842 void generic_fill_statx_attr(struct inode *inode, struct kstat *stat); 2843 extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int); 2844 extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int); 2845 void __inode_add_bytes(struct inode *inode, loff_t bytes); 2846 void inode_add_bytes(struct inode *inode, loff_t bytes); 2847 void __inode_sub_bytes(struct inode *inode, loff_t bytes); 2848 void inode_sub_bytes(struct inode *inode, loff_t bytes); 2849 static inline loff_t __inode_get_bytes(struct inode *inode) 2850 { 2851 return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes; 2852 } 2853 loff_t inode_get_bytes(struct inode *inode); 2854 void inode_set_bytes(struct inode *inode, loff_t bytes); 2855 const char *simple_get_link(struct dentry *, struct inode *, 2856 struct delayed_call *); 2857 extern const struct inode_operations simple_symlink_inode_operations; 2858 2859 extern int iterate_dir(struct file *, struct dir_context *); 2860 2861 int vfs_fstatat(int dfd, const char __user *filename, struct kstat *stat, 2862 int flags); 2863 int vfs_fstat(int fd, struct kstat *stat); 2864 2865 static inline int vfs_stat(const char __user *filename, struct kstat *stat) 2866 { 2867 return vfs_fstatat(AT_FDCWD, filename, stat, 0); 2868 } 2869 static inline int vfs_lstat(const char __user *name, struct kstat *stat) 2870 { 2871 return vfs_fstatat(AT_FDCWD, name, stat, AT_SYMLINK_NOFOLLOW); 2872 } 2873 2874 extern const char *vfs_get_link(struct dentry *, struct delayed_call *); 2875 extern int vfs_readlink(struct dentry *, char __user *, int); 2876 2877 extern struct file_system_type *get_filesystem(struct file_system_type *fs); 2878 extern void put_filesystem(struct file_system_type *fs); 2879 extern struct file_system_type *get_fs_type(const char *name); 2880 extern struct super_block *get_super(struct block_device *); 2881 extern struct super_block *get_active_super(struct block_device *bdev); 2882 extern void drop_super(struct super_block *sb); 2883 extern void drop_super_exclusive(struct super_block *sb); 2884 extern void iterate_supers(void (*)(struct super_block *, void *), void *); 2885 extern void iterate_supers_type(struct file_system_type *, 2886 void (*)(struct super_block *, void *), void *); 2887 2888 extern int dcache_dir_open(struct inode *, struct file *); 2889 extern int dcache_dir_close(struct inode *, struct file *); 2890 extern loff_t dcache_dir_lseek(struct file *, loff_t, int); 2891 extern int dcache_readdir(struct file *, struct dir_context *); 2892 extern int simple_setattr(struct mnt_idmap *, struct dentry *, 2893 struct iattr *); 2894 extern int simple_getattr(struct mnt_idmap *, const struct path *, 2895 struct kstat *, u32, unsigned int); 2896 extern int simple_statfs(struct dentry *, struct kstatfs *); 2897 extern int simple_open(struct inode *inode, struct file *file); 2898 extern int simple_link(struct dentry *, struct inode *, struct dentry *); 2899 extern int simple_unlink(struct inode *, struct dentry *); 2900 extern int simple_rmdir(struct inode *, struct dentry *); 2901 extern int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry, 2902 struct inode *new_dir, struct dentry *new_dentry); 2903 extern int simple_rename(struct mnt_idmap *, struct inode *, 2904 struct dentry *, struct inode *, struct dentry *, 2905 unsigned int); 2906 extern void simple_recursive_removal(struct dentry *, 2907 void (*callback)(struct dentry *)); 2908 extern int noop_fsync(struct file *, loff_t, loff_t, int); 2909 extern ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter); 2910 extern int simple_empty(struct dentry *); 2911 extern int simple_write_begin(struct file *file, struct address_space *mapping, 2912 loff_t pos, unsigned len, 2913 struct page **pagep, void **fsdata); 2914 extern const struct address_space_operations ram_aops; 2915 extern int always_delete_dentry(const struct dentry *); 2916 extern struct inode *alloc_anon_inode(struct super_block *); 2917 extern int simple_nosetlease(struct file *, long, struct file_lock **, void **); 2918 extern const struct dentry_operations simple_dentry_operations; 2919 2920 extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags); 2921 extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *); 2922 extern const struct file_operations simple_dir_operations; 2923 extern const struct inode_operations simple_dir_inode_operations; 2924 extern void make_empty_dir_inode(struct inode *inode); 2925 extern bool is_empty_dir_inode(struct inode *inode); 2926 struct tree_descr { const char *name; const struct file_operations *ops; int mode; }; 2927 struct dentry *d_alloc_name(struct dentry *, const char *); 2928 extern int simple_fill_super(struct super_block *, unsigned long, 2929 const struct tree_descr *); 2930 extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count); 2931 extern void simple_release_fs(struct vfsmount **mount, int *count); 2932 2933 extern ssize_t simple_read_from_buffer(void __user *to, size_t count, 2934 loff_t *ppos, const void *from, size_t available); 2935 extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos, 2936 const void __user *from, size_t count); 2937 2938 extern int __generic_file_fsync(struct file *, loff_t, loff_t, int); 2939 extern int generic_file_fsync(struct file *, loff_t, loff_t, int); 2940 2941 extern int generic_check_addressable(unsigned, u64); 2942 2943 extern void generic_set_encrypted_ci_d_ops(struct dentry *dentry); 2944 2945 int may_setattr(struct mnt_idmap *idmap, struct inode *inode, 2946 unsigned int ia_valid); 2947 int setattr_prepare(struct mnt_idmap *, struct dentry *, struct iattr *); 2948 extern int inode_newsize_ok(const struct inode *, loff_t offset); 2949 void setattr_copy(struct mnt_idmap *, struct inode *inode, 2950 const struct iattr *attr); 2951 2952 extern int file_update_time(struct file *file); 2953 2954 static inline bool vma_is_dax(const struct vm_area_struct *vma) 2955 { 2956 return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host); 2957 } 2958 2959 static inline bool vma_is_fsdax(struct vm_area_struct *vma) 2960 { 2961 struct inode *inode; 2962 2963 if (!IS_ENABLED(CONFIG_FS_DAX) || !vma->vm_file) 2964 return false; 2965 if (!vma_is_dax(vma)) 2966 return false; 2967 inode = file_inode(vma->vm_file); 2968 if (S_ISCHR(inode->i_mode)) 2969 return false; /* device-dax */ 2970 return true; 2971 } 2972 2973 static inline int iocb_flags(struct file *file) 2974 { 2975 int res = 0; 2976 if (file->f_flags & O_APPEND) 2977 res |= IOCB_APPEND; 2978 if (file->f_flags & O_DIRECT) 2979 res |= IOCB_DIRECT; 2980 if (file->f_flags & O_DSYNC) 2981 res |= IOCB_DSYNC; 2982 if (file->f_flags & __O_SYNC) 2983 res |= IOCB_SYNC; 2984 return res; 2985 } 2986 2987 static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags) 2988 { 2989 int kiocb_flags = 0; 2990 2991 /* make sure there's no overlap between RWF and private IOCB flags */ 2992 BUILD_BUG_ON((__force int) RWF_SUPPORTED & IOCB_EVENTFD); 2993 2994 if (!flags) 2995 return 0; 2996 if (unlikely(flags & ~RWF_SUPPORTED)) 2997 return -EOPNOTSUPP; 2998 2999 if (flags & RWF_NOWAIT) { 3000 if (!(ki->ki_filp->f_mode & FMODE_NOWAIT)) 3001 return -EOPNOTSUPP; 3002 kiocb_flags |= IOCB_NOIO; 3003 } 3004 kiocb_flags |= (__force int) (flags & RWF_SUPPORTED); 3005 if (flags & RWF_SYNC) 3006 kiocb_flags |= IOCB_DSYNC; 3007 3008 ki->ki_flags |= kiocb_flags; 3009 return 0; 3010 } 3011 3012 static inline ino_t parent_ino(struct dentry *dentry) 3013 { 3014 ino_t res; 3015 3016 /* 3017 * Don't strictly need d_lock here? If the parent ino could change 3018 * then surely we'd have a deeper race in the caller? 3019 */ 3020 spin_lock(&dentry->d_lock); 3021 res = dentry->d_parent->d_inode->i_ino; 3022 spin_unlock(&dentry->d_lock); 3023 return res; 3024 } 3025 3026 /* Transaction based IO helpers */ 3027 3028 /* 3029 * An argresp is stored in an allocated page and holds the 3030 * size of the argument or response, along with its content 3031 */ 3032 struct simple_transaction_argresp { 3033 ssize_t size; 3034 char data[]; 3035 }; 3036 3037 #define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp)) 3038 3039 char *simple_transaction_get(struct file *file, const char __user *buf, 3040 size_t size); 3041 ssize_t simple_transaction_read(struct file *file, char __user *buf, 3042 size_t size, loff_t *pos); 3043 int simple_transaction_release(struct inode *inode, struct file *file); 3044 3045 void simple_transaction_set(struct file *file, size_t n); 3046 3047 /* 3048 * simple attribute files 3049 * 3050 * These attributes behave similar to those in sysfs: 3051 * 3052 * Writing to an attribute immediately sets a value, an open file can be 3053 * written to multiple times. 3054 * 3055 * Reading from an attribute creates a buffer from the value that might get 3056 * read with multiple read calls. When the attribute has been read 3057 * completely, no further read calls are possible until the file is opened 3058 * again. 3059 * 3060 * All attributes contain a text representation of a numeric value 3061 * that are accessed with the get() and set() functions. 3062 */ 3063 #define DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, __is_signed) \ 3064 static int __fops ## _open(struct inode *inode, struct file *file) \ 3065 { \ 3066 __simple_attr_check_format(__fmt, 0ull); \ 3067 return simple_attr_open(inode, file, __get, __set, __fmt); \ 3068 } \ 3069 static const struct file_operations __fops = { \ 3070 .owner = THIS_MODULE, \ 3071 .open = __fops ## _open, \ 3072 .release = simple_attr_release, \ 3073 .read = simple_attr_read, \ 3074 .write = (__is_signed) ? simple_attr_write_signed : simple_attr_write, \ 3075 .llseek = generic_file_llseek, \ 3076 } 3077 3078 #define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt) \ 3079 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, false) 3080 3081 #define DEFINE_SIMPLE_ATTRIBUTE_SIGNED(__fops, __get, __set, __fmt) \ 3082 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, true) 3083 3084 static inline __printf(1, 2) 3085 void __simple_attr_check_format(const char *fmt, ...) 3086 { 3087 /* don't do anything, just let the compiler check the arguments; */ 3088 } 3089 3090 int simple_attr_open(struct inode *inode, struct file *file, 3091 int (*get)(void *, u64 *), int (*set)(void *, u64), 3092 const char *fmt); 3093 int simple_attr_release(struct inode *inode, struct file *file); 3094 ssize_t simple_attr_read(struct file *file, char __user *buf, 3095 size_t len, loff_t *ppos); 3096 ssize_t simple_attr_write(struct file *file, const char __user *buf, 3097 size_t len, loff_t *ppos); 3098 ssize_t simple_attr_write_signed(struct file *file, const char __user *buf, 3099 size_t len, loff_t *ppos); 3100 3101 struct ctl_table; 3102 int __init list_bdev_fs_names(char *buf, size_t size); 3103 3104 #define __FMODE_EXEC ((__force int) FMODE_EXEC) 3105 #define __FMODE_NONOTIFY ((__force int) FMODE_NONOTIFY) 3106 3107 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE]) 3108 #define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \ 3109 (flag & __FMODE_NONOTIFY))) 3110 3111 static inline bool is_sxid(umode_t mode) 3112 { 3113 return mode & (S_ISUID | S_ISGID); 3114 } 3115 3116 static inline int check_sticky(struct mnt_idmap *idmap, 3117 struct inode *dir, struct inode *inode) 3118 { 3119 if (!(dir->i_mode & S_ISVTX)) 3120 return 0; 3121 3122 return __check_sticky(idmap, dir, inode); 3123 } 3124 3125 static inline void inode_has_no_xattr(struct inode *inode) 3126 { 3127 if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC)) 3128 inode->i_flags |= S_NOSEC; 3129 } 3130 3131 static inline bool is_root_inode(struct inode *inode) 3132 { 3133 return inode == inode->i_sb->s_root->d_inode; 3134 } 3135 3136 static inline bool dir_emit(struct dir_context *ctx, 3137 const char *name, int namelen, 3138 u64 ino, unsigned type) 3139 { 3140 return ctx->actor(ctx, name, namelen, ctx->pos, ino, type); 3141 } 3142 static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx) 3143 { 3144 return ctx->actor(ctx, ".", 1, ctx->pos, 3145 file->f_path.dentry->d_inode->i_ino, DT_DIR); 3146 } 3147 static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx) 3148 { 3149 return ctx->actor(ctx, "..", 2, ctx->pos, 3150 parent_ino(file->f_path.dentry), DT_DIR); 3151 } 3152 static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx) 3153 { 3154 if (ctx->pos == 0) { 3155 if (!dir_emit_dot(file, ctx)) 3156 return false; 3157 ctx->pos = 1; 3158 } 3159 if (ctx->pos == 1) { 3160 if (!dir_emit_dotdot(file, ctx)) 3161 return false; 3162 ctx->pos = 2; 3163 } 3164 return true; 3165 } 3166 static inline bool dir_relax(struct inode *inode) 3167 { 3168 inode_unlock(inode); 3169 inode_lock(inode); 3170 return !IS_DEADDIR(inode); 3171 } 3172 3173 static inline bool dir_relax_shared(struct inode *inode) 3174 { 3175 inode_unlock_shared(inode); 3176 inode_lock_shared(inode); 3177 return !IS_DEADDIR(inode); 3178 } 3179 3180 extern bool path_noexec(const struct path *path); 3181 extern void inode_nohighmem(struct inode *inode); 3182 3183 /* mm/fadvise.c */ 3184 extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len, 3185 int advice); 3186 extern int generic_fadvise(struct file *file, loff_t offset, loff_t len, 3187 int advice); 3188 3189 #endif /* _LINUX_FS_H */ 3190