1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _LINUX_FS_H 3 #define _LINUX_FS_H 4 5 #include <linux/linkage.h> 6 #include <linux/wait_bit.h> 7 #include <linux/kdev_t.h> 8 #include <linux/dcache.h> 9 #include <linux/path.h> 10 #include <linux/stat.h> 11 #include <linux/cache.h> 12 #include <linux/list.h> 13 #include <linux/list_lru.h> 14 #include <linux/llist.h> 15 #include <linux/radix-tree.h> 16 #include <linux/xarray.h> 17 #include <linux/rbtree.h> 18 #include <linux/init.h> 19 #include <linux/pid.h> 20 #include <linux/bug.h> 21 #include <linux/mutex.h> 22 #include <linux/rwsem.h> 23 #include <linux/mm_types.h> 24 #include <linux/capability.h> 25 #include <linux/semaphore.h> 26 #include <linux/fcntl.h> 27 #include <linux/rculist_bl.h> 28 #include <linux/atomic.h> 29 #include <linux/shrinker.h> 30 #include <linux/migrate_mode.h> 31 #include <linux/uidgid.h> 32 #include <linux/lockdep.h> 33 #include <linux/percpu-rwsem.h> 34 #include <linux/workqueue.h> 35 #include <linux/delayed_call.h> 36 #include <linux/uuid.h> 37 #include <linux/errseq.h> 38 #include <linux/ioprio.h> 39 #include <linux/fs_types.h> 40 #include <linux/build_bug.h> 41 #include <linux/stddef.h> 42 #include <linux/mount.h> 43 #include <linux/cred.h> 44 #include <linux/mnt_idmapping.h> 45 #include <linux/slab.h> 46 47 #include <asm/byteorder.h> 48 #include <uapi/linux/fs.h> 49 50 struct backing_dev_info; 51 struct bdi_writeback; 52 struct bio; 53 struct io_comp_batch; 54 struct export_operations; 55 struct fiemap_extent_info; 56 struct hd_geometry; 57 struct iovec; 58 struct kiocb; 59 struct kobject; 60 struct pipe_inode_info; 61 struct poll_table_struct; 62 struct kstatfs; 63 struct vm_area_struct; 64 struct vfsmount; 65 struct cred; 66 struct swap_info_struct; 67 struct seq_file; 68 struct workqueue_struct; 69 struct iov_iter; 70 struct fscrypt_info; 71 struct fscrypt_operations; 72 struct fsverity_info; 73 struct fsverity_operations; 74 struct fs_context; 75 struct fs_parameter_spec; 76 struct fileattr; 77 struct iomap_ops; 78 79 extern void __init inode_init(void); 80 extern void __init inode_init_early(void); 81 extern void __init files_init(void); 82 extern void __init files_maxfiles_init(void); 83 84 extern unsigned long get_max_files(void); 85 extern unsigned int sysctl_nr_open; 86 87 typedef __kernel_rwf_t rwf_t; 88 89 struct buffer_head; 90 typedef int (get_block_t)(struct inode *inode, sector_t iblock, 91 struct buffer_head *bh_result, int create); 92 typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset, 93 ssize_t bytes, void *private); 94 95 #define MAY_EXEC 0x00000001 96 #define MAY_WRITE 0x00000002 97 #define MAY_READ 0x00000004 98 #define MAY_APPEND 0x00000008 99 #define MAY_ACCESS 0x00000010 100 #define MAY_OPEN 0x00000020 101 #define MAY_CHDIR 0x00000040 102 /* called from RCU mode, don't block */ 103 #define MAY_NOT_BLOCK 0x00000080 104 105 /* 106 * flags in file.f_mode. Note that FMODE_READ and FMODE_WRITE must correspond 107 * to O_WRONLY and O_RDWR via the strange trick in do_dentry_open() 108 */ 109 110 /* file is open for reading */ 111 #define FMODE_READ ((__force fmode_t)0x1) 112 /* file is open for writing */ 113 #define FMODE_WRITE ((__force fmode_t)0x2) 114 /* file is seekable */ 115 #define FMODE_LSEEK ((__force fmode_t)0x4) 116 /* file can be accessed using pread */ 117 #define FMODE_PREAD ((__force fmode_t)0x8) 118 /* file can be accessed using pwrite */ 119 #define FMODE_PWRITE ((__force fmode_t)0x10) 120 /* File is opened for execution with sys_execve / sys_uselib */ 121 #define FMODE_EXEC ((__force fmode_t)0x20) 122 /* 32bit hashes as llseek() offset (for directories) */ 123 #define FMODE_32BITHASH ((__force fmode_t)0x200) 124 /* 64bit hashes as llseek() offset (for directories) */ 125 #define FMODE_64BITHASH ((__force fmode_t)0x400) 126 127 /* 128 * Don't update ctime and mtime. 129 * 130 * Currently a special hack for the XFS open_by_handle ioctl, but we'll 131 * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon. 132 */ 133 #define FMODE_NOCMTIME ((__force fmode_t)0x800) 134 135 /* Expect random access pattern */ 136 #define FMODE_RANDOM ((__force fmode_t)0x1000) 137 138 /* File is huge (eg. /dev/mem): treat loff_t as unsigned */ 139 #define FMODE_UNSIGNED_OFFSET ((__force fmode_t)0x2000) 140 141 /* File is opened with O_PATH; almost nothing can be done with it */ 142 #define FMODE_PATH ((__force fmode_t)0x4000) 143 144 /* File needs atomic accesses to f_pos */ 145 #define FMODE_ATOMIC_POS ((__force fmode_t)0x8000) 146 /* Write access to underlying fs */ 147 #define FMODE_WRITER ((__force fmode_t)0x10000) 148 /* Has read method(s) */ 149 #define FMODE_CAN_READ ((__force fmode_t)0x20000) 150 /* Has write method(s) */ 151 #define FMODE_CAN_WRITE ((__force fmode_t)0x40000) 152 153 #define FMODE_OPENED ((__force fmode_t)0x80000) 154 #define FMODE_CREATED ((__force fmode_t)0x100000) 155 156 /* File is stream-like */ 157 #define FMODE_STREAM ((__force fmode_t)0x200000) 158 159 /* File supports DIRECT IO */ 160 #define FMODE_CAN_ODIRECT ((__force fmode_t)0x400000) 161 162 #define FMODE_NOREUSE ((__force fmode_t)0x800000) 163 164 /* File supports non-exclusive O_DIRECT writes from multiple threads */ 165 #define FMODE_DIO_PARALLEL_WRITE ((__force fmode_t)0x1000000) 166 167 /* File is embedded in backing_file object */ 168 #define FMODE_BACKING ((__force fmode_t)0x2000000) 169 170 /* File was opened by fanotify and shouldn't generate fanotify events */ 171 #define FMODE_NONOTIFY ((__force fmode_t)0x4000000) 172 173 /* File is capable of returning -EAGAIN if I/O will block */ 174 #define FMODE_NOWAIT ((__force fmode_t)0x8000000) 175 176 /* File represents mount that needs unmounting */ 177 #define FMODE_NEED_UNMOUNT ((__force fmode_t)0x10000000) 178 179 /* File does not contribute to nr_files count */ 180 #define FMODE_NOACCOUNT ((__force fmode_t)0x20000000) 181 182 /* File supports async buffered reads */ 183 #define FMODE_BUF_RASYNC ((__force fmode_t)0x40000000) 184 185 /* File supports async nowait buffered writes */ 186 #define FMODE_BUF_WASYNC ((__force fmode_t)0x80000000) 187 188 /* 189 * Attribute flags. These should be or-ed together to figure out what 190 * has been changed! 191 */ 192 #define ATTR_MODE (1 << 0) 193 #define ATTR_UID (1 << 1) 194 #define ATTR_GID (1 << 2) 195 #define ATTR_SIZE (1 << 3) 196 #define ATTR_ATIME (1 << 4) 197 #define ATTR_MTIME (1 << 5) 198 #define ATTR_CTIME (1 << 6) 199 #define ATTR_ATIME_SET (1 << 7) 200 #define ATTR_MTIME_SET (1 << 8) 201 #define ATTR_FORCE (1 << 9) /* Not a change, but a change it */ 202 #define ATTR_KILL_SUID (1 << 11) 203 #define ATTR_KILL_SGID (1 << 12) 204 #define ATTR_FILE (1 << 13) 205 #define ATTR_KILL_PRIV (1 << 14) 206 #define ATTR_OPEN (1 << 15) /* Truncating from open(O_TRUNC) */ 207 #define ATTR_TIMES_SET (1 << 16) 208 #define ATTR_TOUCH (1 << 17) 209 210 /* 211 * Whiteout is represented by a char device. The following constants define the 212 * mode and device number to use. 213 */ 214 #define WHITEOUT_MODE 0 215 #define WHITEOUT_DEV 0 216 217 /* 218 * This is the Inode Attributes structure, used for notify_change(). It 219 * uses the above definitions as flags, to know which values have changed. 220 * Also, in this manner, a Filesystem can look at only the values it cares 221 * about. Basically, these are the attributes that the VFS layer can 222 * request to change from the FS layer. 223 * 224 * Derek Atkins <warlord@MIT.EDU> 94-10-20 225 */ 226 struct iattr { 227 unsigned int ia_valid; 228 umode_t ia_mode; 229 /* 230 * The two anonymous unions wrap structures with the same member. 231 * 232 * Filesystems raising FS_ALLOW_IDMAP need to use ia_vfs{g,u}id which 233 * are a dedicated type requiring the filesystem to use the dedicated 234 * helpers. Other filesystem can continue to use ia_{g,u}id until they 235 * have been ported. 236 * 237 * They always contain the same value. In other words FS_ALLOW_IDMAP 238 * pass down the same value on idmapped mounts as they would on regular 239 * mounts. 240 */ 241 union { 242 kuid_t ia_uid; 243 vfsuid_t ia_vfsuid; 244 }; 245 union { 246 kgid_t ia_gid; 247 vfsgid_t ia_vfsgid; 248 }; 249 loff_t ia_size; 250 struct timespec64 ia_atime; 251 struct timespec64 ia_mtime; 252 struct timespec64 ia_ctime; 253 254 /* 255 * Not an attribute, but an auxiliary info for filesystems wanting to 256 * implement an ftruncate() like method. NOTE: filesystem should 257 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL). 258 */ 259 struct file *ia_file; 260 }; 261 262 /* 263 * Includes for diskquotas. 264 */ 265 #include <linux/quota.h> 266 267 /* 268 * Maximum number of layers of fs stack. Needs to be limited to 269 * prevent kernel stack overflow 270 */ 271 #define FILESYSTEM_MAX_STACK_DEPTH 2 272 273 /** 274 * enum positive_aop_returns - aop return codes with specific semantics 275 * 276 * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has 277 * completed, that the page is still locked, and 278 * should be considered active. The VM uses this hint 279 * to return the page to the active list -- it won't 280 * be a candidate for writeback again in the near 281 * future. Other callers must be careful to unlock 282 * the page if they get this return. Returned by 283 * writepage(); 284 * 285 * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has 286 * unlocked it and the page might have been truncated. 287 * The caller should back up to acquiring a new page and 288 * trying again. The aop will be taking reasonable 289 * precautions not to livelock. If the caller held a page 290 * reference, it should drop it before retrying. Returned 291 * by read_folio(). 292 * 293 * address_space_operation functions return these large constants to indicate 294 * special semantics to the caller. These are much larger than the bytes in a 295 * page to allow for functions that return the number of bytes operated on in a 296 * given page. 297 */ 298 299 enum positive_aop_returns { 300 AOP_WRITEPAGE_ACTIVATE = 0x80000, 301 AOP_TRUNCATED_PAGE = 0x80001, 302 }; 303 304 /* 305 * oh the beauties of C type declarations. 306 */ 307 struct page; 308 struct address_space; 309 struct writeback_control; 310 struct readahead_control; 311 312 /* 313 * Write life time hint values. 314 * Stored in struct inode as u8. 315 */ 316 enum rw_hint { 317 WRITE_LIFE_NOT_SET = 0, 318 WRITE_LIFE_NONE = RWH_WRITE_LIFE_NONE, 319 WRITE_LIFE_SHORT = RWH_WRITE_LIFE_SHORT, 320 WRITE_LIFE_MEDIUM = RWH_WRITE_LIFE_MEDIUM, 321 WRITE_LIFE_LONG = RWH_WRITE_LIFE_LONG, 322 WRITE_LIFE_EXTREME = RWH_WRITE_LIFE_EXTREME, 323 }; 324 325 /* Match RWF_* bits to IOCB bits */ 326 #define IOCB_HIPRI (__force int) RWF_HIPRI 327 #define IOCB_DSYNC (__force int) RWF_DSYNC 328 #define IOCB_SYNC (__force int) RWF_SYNC 329 #define IOCB_NOWAIT (__force int) RWF_NOWAIT 330 #define IOCB_APPEND (__force int) RWF_APPEND 331 332 /* non-RWF related bits - start at 16 */ 333 #define IOCB_EVENTFD (1 << 16) 334 #define IOCB_DIRECT (1 << 17) 335 #define IOCB_WRITE (1 << 18) 336 /* iocb->ki_waitq is valid */ 337 #define IOCB_WAITQ (1 << 19) 338 #define IOCB_NOIO (1 << 20) 339 /* can use bio alloc cache */ 340 #define IOCB_ALLOC_CACHE (1 << 21) 341 /* 342 * IOCB_DIO_CALLER_COMP can be set by the iocb owner, to indicate that the 343 * iocb completion can be passed back to the owner for execution from a safe 344 * context rather than needing to be punted through a workqueue. If this 345 * flag is set, the bio completion handling may set iocb->dio_complete to a 346 * handler function and iocb->private to context information for that handler. 347 * The issuer should call the handler with that context information from task 348 * context to complete the processing of the iocb. Note that while this 349 * provides a task context for the dio_complete() callback, it should only be 350 * used on the completion side for non-IO generating completions. It's fine to 351 * call blocking functions from this callback, but they should not wait for 352 * unrelated IO (like cache flushing, new IO generation, etc). 353 */ 354 #define IOCB_DIO_CALLER_COMP (1 << 22) 355 /* kiocb is a read or write operation submitted by fs/aio.c. */ 356 #define IOCB_AIO_RW (1 << 23) 357 358 /* for use in trace events */ 359 #define TRACE_IOCB_STRINGS \ 360 { IOCB_HIPRI, "HIPRI" }, \ 361 { IOCB_DSYNC, "DSYNC" }, \ 362 { IOCB_SYNC, "SYNC" }, \ 363 { IOCB_NOWAIT, "NOWAIT" }, \ 364 { IOCB_APPEND, "APPEND" }, \ 365 { IOCB_EVENTFD, "EVENTFD"}, \ 366 { IOCB_DIRECT, "DIRECT" }, \ 367 { IOCB_WRITE, "WRITE" }, \ 368 { IOCB_WAITQ, "WAITQ" }, \ 369 { IOCB_NOIO, "NOIO" }, \ 370 { IOCB_ALLOC_CACHE, "ALLOC_CACHE" }, \ 371 { IOCB_DIO_CALLER_COMP, "CALLER_COMP" } 372 373 struct kiocb { 374 struct file *ki_filp; 375 loff_t ki_pos; 376 void (*ki_complete)(struct kiocb *iocb, long ret); 377 void *private; 378 int ki_flags; 379 u16 ki_ioprio; /* See linux/ioprio.h */ 380 union { 381 /* 382 * Only used for async buffered reads, where it denotes the 383 * page waitqueue associated with completing the read. Valid 384 * IFF IOCB_WAITQ is set. 385 */ 386 struct wait_page_queue *ki_waitq; 387 /* 388 * Can be used for O_DIRECT IO, where the completion handling 389 * is punted back to the issuer of the IO. May only be set 390 * if IOCB_DIO_CALLER_COMP is set by the issuer, and the issuer 391 * must then check for presence of this handler when ki_complete 392 * is invoked. The data passed in to this handler must be 393 * assigned to ->private when dio_complete is assigned. 394 */ 395 ssize_t (*dio_complete)(void *data); 396 }; 397 }; 398 399 static inline bool is_sync_kiocb(struct kiocb *kiocb) 400 { 401 return kiocb->ki_complete == NULL; 402 } 403 404 struct address_space_operations { 405 int (*writepage)(struct page *page, struct writeback_control *wbc); 406 int (*read_folio)(struct file *, struct folio *); 407 408 /* Write back some dirty pages from this mapping. */ 409 int (*writepages)(struct address_space *, struct writeback_control *); 410 411 /* Mark a folio dirty. Return true if this dirtied it */ 412 bool (*dirty_folio)(struct address_space *, struct folio *); 413 414 void (*readahead)(struct readahead_control *); 415 416 int (*write_begin)(struct file *, struct address_space *mapping, 417 loff_t pos, unsigned len, 418 struct page **pagep, void **fsdata); 419 int (*write_end)(struct file *, struct address_space *mapping, 420 loff_t pos, unsigned len, unsigned copied, 421 struct page *page, void *fsdata); 422 423 /* Unfortunately this kludge is needed for FIBMAP. Don't use it */ 424 sector_t (*bmap)(struct address_space *, sector_t); 425 void (*invalidate_folio) (struct folio *, size_t offset, size_t len); 426 bool (*release_folio)(struct folio *, gfp_t); 427 void (*free_folio)(struct folio *folio); 428 ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter); 429 /* 430 * migrate the contents of a folio to the specified target. If 431 * migrate_mode is MIGRATE_ASYNC, it must not block. 432 */ 433 int (*migrate_folio)(struct address_space *, struct folio *dst, 434 struct folio *src, enum migrate_mode); 435 int (*launder_folio)(struct folio *); 436 bool (*is_partially_uptodate) (struct folio *, size_t from, 437 size_t count); 438 void (*is_dirty_writeback) (struct folio *, bool *dirty, bool *wb); 439 int (*error_remove_page)(struct address_space *, struct page *); 440 441 /* swapfile support */ 442 int (*swap_activate)(struct swap_info_struct *sis, struct file *file, 443 sector_t *span); 444 void (*swap_deactivate)(struct file *file); 445 int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter); 446 }; 447 448 extern const struct address_space_operations empty_aops; 449 450 /** 451 * struct address_space - Contents of a cacheable, mappable object. 452 * @host: Owner, either the inode or the block_device. 453 * @i_pages: Cached pages. 454 * @invalidate_lock: Guards coherency between page cache contents and 455 * file offset->disk block mappings in the filesystem during invalidates. 456 * It is also used to block modification of page cache contents through 457 * memory mappings. 458 * @gfp_mask: Memory allocation flags to use for allocating pages. 459 * @i_mmap_writable: Number of VM_SHARED mappings. 460 * @nr_thps: Number of THPs in the pagecache (non-shmem only). 461 * @i_mmap: Tree of private and shared mappings. 462 * @i_mmap_rwsem: Protects @i_mmap and @i_mmap_writable. 463 * @nrpages: Number of page entries, protected by the i_pages lock. 464 * @writeback_index: Writeback starts here. 465 * @a_ops: Methods. 466 * @flags: Error bits and flags (AS_*). 467 * @wb_err: The most recent error which has occurred. 468 * @private_lock: For use by the owner of the address_space. 469 * @private_list: For use by the owner of the address_space. 470 * @private_data: For use by the owner of the address_space. 471 */ 472 struct address_space { 473 struct inode *host; 474 struct xarray i_pages; 475 struct rw_semaphore invalidate_lock; 476 gfp_t gfp_mask; 477 atomic_t i_mmap_writable; 478 #ifdef CONFIG_READ_ONLY_THP_FOR_FS 479 /* number of thp, only for non-shmem files */ 480 atomic_t nr_thps; 481 #endif 482 struct rb_root_cached i_mmap; 483 unsigned long nrpages; 484 pgoff_t writeback_index; 485 const struct address_space_operations *a_ops; 486 unsigned long flags; 487 struct rw_semaphore i_mmap_rwsem; 488 errseq_t wb_err; 489 spinlock_t private_lock; 490 struct list_head private_list; 491 void *private_data; 492 } __attribute__((aligned(sizeof(long)))) __randomize_layout; 493 /* 494 * On most architectures that alignment is already the case; but 495 * must be enforced here for CRIS, to let the least significant bit 496 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON. 497 */ 498 499 /* XArray tags, for tagging dirty and writeback pages in the pagecache. */ 500 #define PAGECACHE_TAG_DIRTY XA_MARK_0 501 #define PAGECACHE_TAG_WRITEBACK XA_MARK_1 502 #define PAGECACHE_TAG_TOWRITE XA_MARK_2 503 504 /* 505 * Returns true if any of the pages in the mapping are marked with the tag. 506 */ 507 static inline bool mapping_tagged(struct address_space *mapping, xa_mark_t tag) 508 { 509 return xa_marked(&mapping->i_pages, tag); 510 } 511 512 static inline void i_mmap_lock_write(struct address_space *mapping) 513 { 514 down_write(&mapping->i_mmap_rwsem); 515 } 516 517 static inline int i_mmap_trylock_write(struct address_space *mapping) 518 { 519 return down_write_trylock(&mapping->i_mmap_rwsem); 520 } 521 522 static inline void i_mmap_unlock_write(struct address_space *mapping) 523 { 524 up_write(&mapping->i_mmap_rwsem); 525 } 526 527 static inline int i_mmap_trylock_read(struct address_space *mapping) 528 { 529 return down_read_trylock(&mapping->i_mmap_rwsem); 530 } 531 532 static inline void i_mmap_lock_read(struct address_space *mapping) 533 { 534 down_read(&mapping->i_mmap_rwsem); 535 } 536 537 static inline void i_mmap_unlock_read(struct address_space *mapping) 538 { 539 up_read(&mapping->i_mmap_rwsem); 540 } 541 542 static inline void i_mmap_assert_locked(struct address_space *mapping) 543 { 544 lockdep_assert_held(&mapping->i_mmap_rwsem); 545 } 546 547 static inline void i_mmap_assert_write_locked(struct address_space *mapping) 548 { 549 lockdep_assert_held_write(&mapping->i_mmap_rwsem); 550 } 551 552 /* 553 * Might pages of this file be mapped into userspace? 554 */ 555 static inline int mapping_mapped(struct address_space *mapping) 556 { 557 return !RB_EMPTY_ROOT(&mapping->i_mmap.rb_root); 558 } 559 560 /* 561 * Might pages of this file have been modified in userspace? 562 * Note that i_mmap_writable counts all VM_SHARED vmas: do_mmap 563 * marks vma as VM_SHARED if it is shared, and the file was opened for 564 * writing i.e. vma may be mprotected writable even if now readonly. 565 * 566 * If i_mmap_writable is negative, no new writable mappings are allowed. You 567 * can only deny writable mappings, if none exists right now. 568 */ 569 static inline int mapping_writably_mapped(struct address_space *mapping) 570 { 571 return atomic_read(&mapping->i_mmap_writable) > 0; 572 } 573 574 static inline int mapping_map_writable(struct address_space *mapping) 575 { 576 return atomic_inc_unless_negative(&mapping->i_mmap_writable) ? 577 0 : -EPERM; 578 } 579 580 static inline void mapping_unmap_writable(struct address_space *mapping) 581 { 582 atomic_dec(&mapping->i_mmap_writable); 583 } 584 585 static inline int mapping_deny_writable(struct address_space *mapping) 586 { 587 return atomic_dec_unless_positive(&mapping->i_mmap_writable) ? 588 0 : -EBUSY; 589 } 590 591 static inline void mapping_allow_writable(struct address_space *mapping) 592 { 593 atomic_inc(&mapping->i_mmap_writable); 594 } 595 596 /* 597 * Use sequence counter to get consistent i_size on 32-bit processors. 598 */ 599 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 600 #include <linux/seqlock.h> 601 #define __NEED_I_SIZE_ORDERED 602 #define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount) 603 #else 604 #define i_size_ordered_init(inode) do { } while (0) 605 #endif 606 607 struct posix_acl; 608 #define ACL_NOT_CACHED ((void *)(-1)) 609 /* 610 * ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to 611 * cache the ACL. This also means that ->get_inode_acl() can be called in RCU 612 * mode with the LOOKUP_RCU flag. 613 */ 614 #define ACL_DONT_CACHE ((void *)(-3)) 615 616 static inline struct posix_acl * 617 uncached_acl_sentinel(struct task_struct *task) 618 { 619 return (void *)task + 1; 620 } 621 622 static inline bool 623 is_uncached_acl(struct posix_acl *acl) 624 { 625 return (long)acl & 1; 626 } 627 628 #define IOP_FASTPERM 0x0001 629 #define IOP_LOOKUP 0x0002 630 #define IOP_NOFOLLOW 0x0004 631 #define IOP_XATTR 0x0008 632 #define IOP_DEFAULT_READLINK 0x0010 633 634 struct fsnotify_mark_connector; 635 636 /* 637 * Keep mostly read-only and often accessed (especially for 638 * the RCU path lookup and 'stat' data) fields at the beginning 639 * of the 'struct inode' 640 */ 641 struct inode { 642 umode_t i_mode; 643 unsigned short i_opflags; 644 kuid_t i_uid; 645 kgid_t i_gid; 646 unsigned int i_flags; 647 648 #ifdef CONFIG_FS_POSIX_ACL 649 struct posix_acl *i_acl; 650 struct posix_acl *i_default_acl; 651 #endif 652 653 const struct inode_operations *i_op; 654 struct super_block *i_sb; 655 struct address_space *i_mapping; 656 657 #ifdef CONFIG_SECURITY 658 void *i_security; 659 #endif 660 661 /* Stat data, not accessed from path walking */ 662 unsigned long i_ino; 663 /* 664 * Filesystems may only read i_nlink directly. They shall use the 665 * following functions for modification: 666 * 667 * (set|clear|inc|drop)_nlink 668 * inode_(inc|dec)_link_count 669 */ 670 union { 671 const unsigned int i_nlink; 672 unsigned int __i_nlink; 673 }; 674 dev_t i_rdev; 675 loff_t i_size; 676 struct timespec64 i_atime; 677 struct timespec64 i_mtime; 678 struct timespec64 __i_ctime; /* use inode_*_ctime accessors! */ 679 spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */ 680 unsigned short i_bytes; 681 u8 i_blkbits; 682 u8 i_write_hint; 683 blkcnt_t i_blocks; 684 685 #ifdef __NEED_I_SIZE_ORDERED 686 seqcount_t i_size_seqcount; 687 #endif 688 689 /* Misc */ 690 unsigned long i_state; 691 struct rw_semaphore i_rwsem; 692 693 unsigned long dirtied_when; /* jiffies of first dirtying */ 694 unsigned long dirtied_time_when; 695 696 struct hlist_node i_hash; 697 struct list_head i_io_list; /* backing dev IO list */ 698 #ifdef CONFIG_CGROUP_WRITEBACK 699 struct bdi_writeback *i_wb; /* the associated cgroup wb */ 700 701 /* foreign inode detection, see wbc_detach_inode() */ 702 int i_wb_frn_winner; 703 u16 i_wb_frn_avg_time; 704 u16 i_wb_frn_history; 705 #endif 706 struct list_head i_lru; /* inode LRU list */ 707 struct list_head i_sb_list; 708 struct list_head i_wb_list; /* backing dev writeback list */ 709 union { 710 struct hlist_head i_dentry; 711 struct rcu_head i_rcu; 712 }; 713 atomic64_t i_version; 714 atomic64_t i_sequence; /* see futex */ 715 atomic_t i_count; 716 atomic_t i_dio_count; 717 atomic_t i_writecount; 718 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING) 719 atomic_t i_readcount; /* struct files open RO */ 720 #endif 721 union { 722 const struct file_operations *i_fop; /* former ->i_op->default_file_ops */ 723 void (*free_inode)(struct inode *); 724 }; 725 struct file_lock_context *i_flctx; 726 struct address_space i_data; 727 struct list_head i_devices; 728 union { 729 struct pipe_inode_info *i_pipe; 730 struct cdev *i_cdev; 731 char *i_link; 732 unsigned i_dir_seq; 733 }; 734 735 __u32 i_generation; 736 737 #ifdef CONFIG_FSNOTIFY 738 __u32 i_fsnotify_mask; /* all events this inode cares about */ 739 struct fsnotify_mark_connector __rcu *i_fsnotify_marks; 740 #endif 741 742 #ifdef CONFIG_FS_ENCRYPTION 743 struct fscrypt_info *i_crypt_info; 744 #endif 745 746 #ifdef CONFIG_FS_VERITY 747 struct fsverity_info *i_verity_info; 748 #endif 749 750 void *i_private; /* fs or device private pointer */ 751 } __randomize_layout; 752 753 struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode); 754 755 static inline unsigned int i_blocksize(const struct inode *node) 756 { 757 return (1 << node->i_blkbits); 758 } 759 760 static inline int inode_unhashed(struct inode *inode) 761 { 762 return hlist_unhashed(&inode->i_hash); 763 } 764 765 /* 766 * __mark_inode_dirty expects inodes to be hashed. Since we don't 767 * want special inodes in the fileset inode space, we make them 768 * appear hashed, but do not put on any lists. hlist_del() 769 * will work fine and require no locking. 770 */ 771 static inline void inode_fake_hash(struct inode *inode) 772 { 773 hlist_add_fake(&inode->i_hash); 774 } 775 776 /* 777 * inode->i_mutex nesting subclasses for the lock validator: 778 * 779 * 0: the object of the current VFS operation 780 * 1: parent 781 * 2: child/target 782 * 3: xattr 783 * 4: second non-directory 784 * 5: second parent (when locking independent directories in rename) 785 * 786 * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two 787 * non-directories at once. 788 * 789 * The locking order between these classes is 790 * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory 791 */ 792 enum inode_i_mutex_lock_class 793 { 794 I_MUTEX_NORMAL, 795 I_MUTEX_PARENT, 796 I_MUTEX_CHILD, 797 I_MUTEX_XATTR, 798 I_MUTEX_NONDIR2, 799 I_MUTEX_PARENT2, 800 }; 801 802 static inline void inode_lock(struct inode *inode) 803 { 804 down_write(&inode->i_rwsem); 805 } 806 807 static inline void inode_unlock(struct inode *inode) 808 { 809 up_write(&inode->i_rwsem); 810 } 811 812 static inline void inode_lock_shared(struct inode *inode) 813 { 814 down_read(&inode->i_rwsem); 815 } 816 817 static inline void inode_unlock_shared(struct inode *inode) 818 { 819 up_read(&inode->i_rwsem); 820 } 821 822 static inline int inode_trylock(struct inode *inode) 823 { 824 return down_write_trylock(&inode->i_rwsem); 825 } 826 827 static inline int inode_trylock_shared(struct inode *inode) 828 { 829 return down_read_trylock(&inode->i_rwsem); 830 } 831 832 static inline int inode_is_locked(struct inode *inode) 833 { 834 return rwsem_is_locked(&inode->i_rwsem); 835 } 836 837 static inline void inode_lock_nested(struct inode *inode, unsigned subclass) 838 { 839 down_write_nested(&inode->i_rwsem, subclass); 840 } 841 842 static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass) 843 { 844 down_read_nested(&inode->i_rwsem, subclass); 845 } 846 847 static inline void filemap_invalidate_lock(struct address_space *mapping) 848 { 849 down_write(&mapping->invalidate_lock); 850 } 851 852 static inline void filemap_invalidate_unlock(struct address_space *mapping) 853 { 854 up_write(&mapping->invalidate_lock); 855 } 856 857 static inline void filemap_invalidate_lock_shared(struct address_space *mapping) 858 { 859 down_read(&mapping->invalidate_lock); 860 } 861 862 static inline int filemap_invalidate_trylock_shared( 863 struct address_space *mapping) 864 { 865 return down_read_trylock(&mapping->invalidate_lock); 866 } 867 868 static inline void filemap_invalidate_unlock_shared( 869 struct address_space *mapping) 870 { 871 up_read(&mapping->invalidate_lock); 872 } 873 874 void lock_two_nondirectories(struct inode *, struct inode*); 875 void unlock_two_nondirectories(struct inode *, struct inode*); 876 877 void filemap_invalidate_lock_two(struct address_space *mapping1, 878 struct address_space *mapping2); 879 void filemap_invalidate_unlock_two(struct address_space *mapping1, 880 struct address_space *mapping2); 881 882 883 /* 884 * NOTE: in a 32bit arch with a preemptable kernel and 885 * an UP compile the i_size_read/write must be atomic 886 * with respect to the local cpu (unlike with preempt disabled), 887 * but they don't need to be atomic with respect to other cpus like in 888 * true SMP (so they need either to either locally disable irq around 889 * the read or for example on x86 they can be still implemented as a 890 * cmpxchg8b without the need of the lock prefix). For SMP compiles 891 * and 64bit archs it makes no difference if preempt is enabled or not. 892 */ 893 static inline loff_t i_size_read(const struct inode *inode) 894 { 895 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 896 loff_t i_size; 897 unsigned int seq; 898 899 do { 900 seq = read_seqcount_begin(&inode->i_size_seqcount); 901 i_size = inode->i_size; 902 } while (read_seqcount_retry(&inode->i_size_seqcount, seq)); 903 return i_size; 904 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION) 905 loff_t i_size; 906 907 preempt_disable(); 908 i_size = inode->i_size; 909 preempt_enable(); 910 return i_size; 911 #else 912 return inode->i_size; 913 #endif 914 } 915 916 /* 917 * NOTE: unlike i_size_read(), i_size_write() does need locking around it 918 * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount 919 * can be lost, resulting in subsequent i_size_read() calls spinning forever. 920 */ 921 static inline void i_size_write(struct inode *inode, loff_t i_size) 922 { 923 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 924 preempt_disable(); 925 write_seqcount_begin(&inode->i_size_seqcount); 926 inode->i_size = i_size; 927 write_seqcount_end(&inode->i_size_seqcount); 928 preempt_enable(); 929 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION) 930 preempt_disable(); 931 inode->i_size = i_size; 932 preempt_enable(); 933 #else 934 inode->i_size = i_size; 935 #endif 936 } 937 938 static inline unsigned iminor(const struct inode *inode) 939 { 940 return MINOR(inode->i_rdev); 941 } 942 943 static inline unsigned imajor(const struct inode *inode) 944 { 945 return MAJOR(inode->i_rdev); 946 } 947 948 struct fown_struct { 949 rwlock_t lock; /* protects pid, uid, euid fields */ 950 struct pid *pid; /* pid or -pgrp where SIGIO should be sent */ 951 enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */ 952 kuid_t uid, euid; /* uid/euid of process setting the owner */ 953 int signum; /* posix.1b rt signal to be delivered on IO */ 954 }; 955 956 /** 957 * struct file_ra_state - Track a file's readahead state. 958 * @start: Where the most recent readahead started. 959 * @size: Number of pages read in the most recent readahead. 960 * @async_size: Numer of pages that were/are not needed immediately 961 * and so were/are genuinely "ahead". Start next readahead when 962 * the first of these pages is accessed. 963 * @ra_pages: Maximum size of a readahead request, copied from the bdi. 964 * @mmap_miss: How many mmap accesses missed in the page cache. 965 * @prev_pos: The last byte in the most recent read request. 966 * 967 * When this structure is passed to ->readahead(), the "most recent" 968 * readahead means the current readahead. 969 */ 970 struct file_ra_state { 971 pgoff_t start; 972 unsigned int size; 973 unsigned int async_size; 974 unsigned int ra_pages; 975 unsigned int mmap_miss; 976 loff_t prev_pos; 977 }; 978 979 /* 980 * Check if @index falls in the readahead windows. 981 */ 982 static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index) 983 { 984 return (index >= ra->start && 985 index < ra->start + ra->size); 986 } 987 988 /* 989 * f_{lock,count,pos_lock} members can be highly contended and share 990 * the same cacheline. f_{lock,mode} are very frequently used together 991 * and so share the same cacheline as well. The read-mostly 992 * f_{path,inode,op} are kept on a separate cacheline. 993 */ 994 struct file { 995 union { 996 struct llist_node f_llist; 997 struct rcu_head f_rcuhead; 998 unsigned int f_iocb_flags; 999 }; 1000 1001 /* 1002 * Protects f_ep, f_flags. 1003 * Must not be taken from IRQ context. 1004 */ 1005 spinlock_t f_lock; 1006 fmode_t f_mode; 1007 atomic_long_t f_count; 1008 struct mutex f_pos_lock; 1009 loff_t f_pos; 1010 unsigned int f_flags; 1011 struct fown_struct f_owner; 1012 const struct cred *f_cred; 1013 struct file_ra_state f_ra; 1014 struct path f_path; 1015 struct inode *f_inode; /* cached value */ 1016 const struct file_operations *f_op; 1017 1018 u64 f_version; 1019 #ifdef CONFIG_SECURITY 1020 void *f_security; 1021 #endif 1022 /* needed for tty driver, and maybe others */ 1023 void *private_data; 1024 1025 #ifdef CONFIG_EPOLL 1026 /* Used by fs/eventpoll.c to link all the hooks to this file */ 1027 struct hlist_head *f_ep; 1028 #endif /* #ifdef CONFIG_EPOLL */ 1029 struct address_space *f_mapping; 1030 errseq_t f_wb_err; 1031 errseq_t f_sb_err; /* for syncfs */ 1032 } __randomize_layout 1033 __attribute__((aligned(4))); /* lest something weird decides that 2 is OK */ 1034 1035 struct file_handle { 1036 __u32 handle_bytes; 1037 int handle_type; 1038 /* file identifier */ 1039 unsigned char f_handle[]; 1040 }; 1041 1042 static inline struct file *get_file(struct file *f) 1043 { 1044 atomic_long_inc(&f->f_count); 1045 return f; 1046 } 1047 #define get_file_rcu(x) atomic_long_inc_not_zero(&(x)->f_count) 1048 #define file_count(x) atomic_long_read(&(x)->f_count) 1049 1050 #define MAX_NON_LFS ((1UL<<31) - 1) 1051 1052 /* Page cache limit. The filesystems should put that into their s_maxbytes 1053 limits, otherwise bad things can happen in VM. */ 1054 #if BITS_PER_LONG==32 1055 #define MAX_LFS_FILESIZE ((loff_t)ULONG_MAX << PAGE_SHIFT) 1056 #elif BITS_PER_LONG==64 1057 #define MAX_LFS_FILESIZE ((loff_t)LLONG_MAX) 1058 #endif 1059 1060 /* legacy typedef, should eventually be removed */ 1061 typedef void *fl_owner_t; 1062 1063 struct file_lock; 1064 1065 /* The following constant reflects the upper bound of the file/locking space */ 1066 #ifndef OFFSET_MAX 1067 #define OFFSET_MAX type_max(loff_t) 1068 #define OFFT_OFFSET_MAX type_max(off_t) 1069 #endif 1070 1071 extern void send_sigio(struct fown_struct *fown, int fd, int band); 1072 1073 static inline struct inode *file_inode(const struct file *f) 1074 { 1075 return f->f_inode; 1076 } 1077 1078 static inline struct dentry *file_dentry(const struct file *file) 1079 { 1080 return d_real(file->f_path.dentry, file_inode(file)); 1081 } 1082 1083 struct fasync_struct { 1084 rwlock_t fa_lock; 1085 int magic; 1086 int fa_fd; 1087 struct fasync_struct *fa_next; /* singly linked list */ 1088 struct file *fa_file; 1089 struct rcu_head fa_rcu; 1090 }; 1091 1092 #define FASYNC_MAGIC 0x4601 1093 1094 /* SMP safe fasync helpers: */ 1095 extern int fasync_helper(int, struct file *, int, struct fasync_struct **); 1096 extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *); 1097 extern int fasync_remove_entry(struct file *, struct fasync_struct **); 1098 extern struct fasync_struct *fasync_alloc(void); 1099 extern void fasync_free(struct fasync_struct *); 1100 1101 /* can be called from interrupts */ 1102 extern void kill_fasync(struct fasync_struct **, int, int); 1103 1104 extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force); 1105 extern int f_setown(struct file *filp, int who, int force); 1106 extern void f_delown(struct file *filp); 1107 extern pid_t f_getown(struct file *filp); 1108 extern int send_sigurg(struct fown_struct *fown); 1109 1110 /* 1111 * sb->s_flags. Note that these mirror the equivalent MS_* flags where 1112 * represented in both. 1113 */ 1114 #define SB_RDONLY BIT(0) /* Mount read-only */ 1115 #define SB_NOSUID BIT(1) /* Ignore suid and sgid bits */ 1116 #define SB_NODEV BIT(2) /* Disallow access to device special files */ 1117 #define SB_NOEXEC BIT(3) /* Disallow program execution */ 1118 #define SB_SYNCHRONOUS BIT(4) /* Writes are synced at once */ 1119 #define SB_MANDLOCK BIT(6) /* Allow mandatory locks on an FS */ 1120 #define SB_DIRSYNC BIT(7) /* Directory modifications are synchronous */ 1121 #define SB_NOATIME BIT(10) /* Do not update access times. */ 1122 #define SB_NODIRATIME BIT(11) /* Do not update directory access times */ 1123 #define SB_SILENT BIT(15) 1124 #define SB_POSIXACL BIT(16) /* VFS does not apply the umask */ 1125 #define SB_INLINECRYPT BIT(17) /* Use blk-crypto for encrypted files */ 1126 #define SB_KERNMOUNT BIT(22) /* this is a kern_mount call */ 1127 #define SB_I_VERSION BIT(23) /* Update inode I_version field */ 1128 #define SB_LAZYTIME BIT(25) /* Update the on-disk [acm]times lazily */ 1129 1130 /* These sb flags are internal to the kernel */ 1131 #define SB_DEAD BIT(21) 1132 #define SB_DYING BIT(24) 1133 #define SB_SUBMOUNT BIT(26) 1134 #define SB_FORCE BIT(27) 1135 #define SB_NOSEC BIT(28) 1136 #define SB_BORN BIT(29) 1137 #define SB_ACTIVE BIT(30) 1138 #define SB_NOUSER BIT(31) 1139 1140 /* These flags relate to encoding and casefolding */ 1141 #define SB_ENC_STRICT_MODE_FL (1 << 0) 1142 1143 #define sb_has_strict_encoding(sb) \ 1144 (sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL) 1145 1146 /* 1147 * Umount options 1148 */ 1149 1150 #define MNT_FORCE 0x00000001 /* Attempt to forcibily umount */ 1151 #define MNT_DETACH 0x00000002 /* Just detach from the tree */ 1152 #define MNT_EXPIRE 0x00000004 /* Mark for expiry */ 1153 #define UMOUNT_NOFOLLOW 0x00000008 /* Don't follow symlink on umount */ 1154 #define UMOUNT_UNUSED 0x80000000 /* Flag guaranteed to be unused */ 1155 1156 /* sb->s_iflags */ 1157 #define SB_I_CGROUPWB 0x00000001 /* cgroup-aware writeback enabled */ 1158 #define SB_I_NOEXEC 0x00000002 /* Ignore executables on this fs */ 1159 #define SB_I_NODEV 0x00000004 /* Ignore devices on this fs */ 1160 #define SB_I_STABLE_WRITES 0x00000008 /* don't modify blks until WB is done */ 1161 1162 /* sb->s_iflags to limit user namespace mounts */ 1163 #define SB_I_USERNS_VISIBLE 0x00000010 /* fstype already mounted */ 1164 #define SB_I_IMA_UNVERIFIABLE_SIGNATURE 0x00000020 1165 #define SB_I_UNTRUSTED_MOUNTER 0x00000040 1166 1167 #define SB_I_SKIP_SYNC 0x00000100 /* Skip superblock at global sync */ 1168 #define SB_I_PERSB_BDI 0x00000200 /* has a per-sb bdi */ 1169 #define SB_I_TS_EXPIRY_WARNED 0x00000400 /* warned about timestamp range expiry */ 1170 #define SB_I_RETIRED 0x00000800 /* superblock shouldn't be reused */ 1171 1172 /* Possible states of 'frozen' field */ 1173 enum { 1174 SB_UNFROZEN = 0, /* FS is unfrozen */ 1175 SB_FREEZE_WRITE = 1, /* Writes, dir ops, ioctls frozen */ 1176 SB_FREEZE_PAGEFAULT = 2, /* Page faults stopped as well */ 1177 SB_FREEZE_FS = 3, /* For internal FS use (e.g. to stop 1178 * internal threads if needed) */ 1179 SB_FREEZE_COMPLETE = 4, /* ->freeze_fs finished successfully */ 1180 }; 1181 1182 #define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1) 1183 1184 struct sb_writers { 1185 unsigned short frozen; /* Is sb frozen? */ 1186 unsigned short freeze_holders; /* Who froze fs? */ 1187 struct percpu_rw_semaphore rw_sem[SB_FREEZE_LEVELS]; 1188 }; 1189 1190 struct super_block { 1191 struct list_head s_list; /* Keep this first */ 1192 dev_t s_dev; /* search index; _not_ kdev_t */ 1193 unsigned char s_blocksize_bits; 1194 unsigned long s_blocksize; 1195 loff_t s_maxbytes; /* Max file size */ 1196 struct file_system_type *s_type; 1197 const struct super_operations *s_op; 1198 const struct dquot_operations *dq_op; 1199 const struct quotactl_ops *s_qcop; 1200 const struct export_operations *s_export_op; 1201 unsigned long s_flags; 1202 unsigned long s_iflags; /* internal SB_I_* flags */ 1203 unsigned long s_magic; 1204 struct dentry *s_root; 1205 struct rw_semaphore s_umount; 1206 int s_count; 1207 atomic_t s_active; 1208 #ifdef CONFIG_SECURITY 1209 void *s_security; 1210 #endif 1211 const struct xattr_handler **s_xattr; 1212 #ifdef CONFIG_FS_ENCRYPTION 1213 const struct fscrypt_operations *s_cop; 1214 struct fscrypt_keyring *s_master_keys; /* master crypto keys in use */ 1215 #endif 1216 #ifdef CONFIG_FS_VERITY 1217 const struct fsverity_operations *s_vop; 1218 #endif 1219 #if IS_ENABLED(CONFIG_UNICODE) 1220 struct unicode_map *s_encoding; 1221 __u16 s_encoding_flags; 1222 #endif 1223 struct hlist_bl_head s_roots; /* alternate root dentries for NFS */ 1224 struct list_head s_mounts; /* list of mounts; _not_ for fs use */ 1225 struct block_device *s_bdev; 1226 struct backing_dev_info *s_bdi; 1227 struct mtd_info *s_mtd; 1228 struct hlist_node s_instances; 1229 unsigned int s_quota_types; /* Bitmask of supported quota types */ 1230 struct quota_info s_dquot; /* Diskquota specific options */ 1231 1232 struct sb_writers s_writers; 1233 1234 /* 1235 * Keep s_fs_info, s_time_gran, s_fsnotify_mask, and 1236 * s_fsnotify_marks together for cache efficiency. They are frequently 1237 * accessed and rarely modified. 1238 */ 1239 void *s_fs_info; /* Filesystem private info */ 1240 1241 /* Granularity of c/m/atime in ns (cannot be worse than a second) */ 1242 u32 s_time_gran; 1243 /* Time limits for c/m/atime in seconds */ 1244 time64_t s_time_min; 1245 time64_t s_time_max; 1246 #ifdef CONFIG_FSNOTIFY 1247 __u32 s_fsnotify_mask; 1248 struct fsnotify_mark_connector __rcu *s_fsnotify_marks; 1249 #endif 1250 1251 char s_id[32]; /* Informational name */ 1252 uuid_t s_uuid; /* UUID */ 1253 1254 unsigned int s_max_links; 1255 1256 /* 1257 * The next field is for VFS *only*. No filesystems have any business 1258 * even looking at it. You had been warned. 1259 */ 1260 struct mutex s_vfs_rename_mutex; /* Kludge */ 1261 1262 /* 1263 * Filesystem subtype. If non-empty the filesystem type field 1264 * in /proc/mounts will be "type.subtype" 1265 */ 1266 const char *s_subtype; 1267 1268 const struct dentry_operations *s_d_op; /* default d_op for dentries */ 1269 1270 struct shrinker s_shrink; /* per-sb shrinker handle */ 1271 1272 /* Number of inodes with nlink == 0 but still referenced */ 1273 atomic_long_t s_remove_count; 1274 1275 /* 1276 * Number of inode/mount/sb objects that are being watched, note that 1277 * inodes objects are currently double-accounted. 1278 */ 1279 atomic_long_t s_fsnotify_connectors; 1280 1281 /* Read-only state of the superblock is being changed */ 1282 int s_readonly_remount; 1283 1284 /* per-sb errseq_t for reporting writeback errors via syncfs */ 1285 errseq_t s_wb_err; 1286 1287 /* AIO completions deferred from interrupt context */ 1288 struct workqueue_struct *s_dio_done_wq; 1289 struct hlist_head s_pins; 1290 1291 /* 1292 * Owning user namespace and default context in which to 1293 * interpret filesystem uids, gids, quotas, device nodes, 1294 * xattrs and security labels. 1295 */ 1296 struct user_namespace *s_user_ns; 1297 1298 /* 1299 * The list_lru structure is essentially just a pointer to a table 1300 * of per-node lru lists, each of which has its own spinlock. 1301 * There is no need to put them into separate cachelines. 1302 */ 1303 struct list_lru s_dentry_lru; 1304 struct list_lru s_inode_lru; 1305 struct rcu_head rcu; 1306 struct work_struct destroy_work; 1307 1308 struct mutex s_sync_lock; /* sync serialisation lock */ 1309 1310 /* 1311 * Indicates how deep in a filesystem stack this SB is 1312 */ 1313 int s_stack_depth; 1314 1315 /* s_inode_list_lock protects s_inodes */ 1316 spinlock_t s_inode_list_lock ____cacheline_aligned_in_smp; 1317 struct list_head s_inodes; /* all inodes */ 1318 1319 spinlock_t s_inode_wblist_lock; 1320 struct list_head s_inodes_wb; /* writeback inodes */ 1321 } __randomize_layout; 1322 1323 static inline struct user_namespace *i_user_ns(const struct inode *inode) 1324 { 1325 return inode->i_sb->s_user_ns; 1326 } 1327 1328 /* Helper functions so that in most cases filesystems will 1329 * not need to deal directly with kuid_t and kgid_t and can 1330 * instead deal with the raw numeric values that are stored 1331 * in the filesystem. 1332 */ 1333 static inline uid_t i_uid_read(const struct inode *inode) 1334 { 1335 return from_kuid(i_user_ns(inode), inode->i_uid); 1336 } 1337 1338 static inline gid_t i_gid_read(const struct inode *inode) 1339 { 1340 return from_kgid(i_user_ns(inode), inode->i_gid); 1341 } 1342 1343 static inline void i_uid_write(struct inode *inode, uid_t uid) 1344 { 1345 inode->i_uid = make_kuid(i_user_ns(inode), uid); 1346 } 1347 1348 static inline void i_gid_write(struct inode *inode, gid_t gid) 1349 { 1350 inode->i_gid = make_kgid(i_user_ns(inode), gid); 1351 } 1352 1353 /** 1354 * i_uid_into_vfsuid - map an inode's i_uid down according to an idmapping 1355 * @idmap: idmap of the mount the inode was found from 1356 * @inode: inode to map 1357 * 1358 * Return: whe inode's i_uid mapped down according to @idmap. 1359 * If the inode's i_uid has no mapping INVALID_VFSUID is returned. 1360 */ 1361 static inline vfsuid_t i_uid_into_vfsuid(struct mnt_idmap *idmap, 1362 const struct inode *inode) 1363 { 1364 return make_vfsuid(idmap, i_user_ns(inode), inode->i_uid); 1365 } 1366 1367 /** 1368 * i_uid_needs_update - check whether inode's i_uid needs to be updated 1369 * @idmap: idmap of the mount the inode was found from 1370 * @attr: the new attributes of @inode 1371 * @inode: the inode to update 1372 * 1373 * Check whether the $inode's i_uid field needs to be updated taking idmapped 1374 * mounts into account if the filesystem supports it. 1375 * 1376 * Return: true if @inode's i_uid field needs to be updated, false if not. 1377 */ 1378 static inline bool i_uid_needs_update(struct mnt_idmap *idmap, 1379 const struct iattr *attr, 1380 const struct inode *inode) 1381 { 1382 return ((attr->ia_valid & ATTR_UID) && 1383 !vfsuid_eq(attr->ia_vfsuid, 1384 i_uid_into_vfsuid(idmap, inode))); 1385 } 1386 1387 /** 1388 * i_uid_update - update @inode's i_uid field 1389 * @idmap: idmap of the mount the inode was found from 1390 * @attr: the new attributes of @inode 1391 * @inode: the inode to update 1392 * 1393 * Safely update @inode's i_uid field translating the vfsuid of any idmapped 1394 * mount into the filesystem kuid. 1395 */ 1396 static inline void i_uid_update(struct mnt_idmap *idmap, 1397 const struct iattr *attr, 1398 struct inode *inode) 1399 { 1400 if (attr->ia_valid & ATTR_UID) 1401 inode->i_uid = from_vfsuid(idmap, i_user_ns(inode), 1402 attr->ia_vfsuid); 1403 } 1404 1405 /** 1406 * i_gid_into_vfsgid - map an inode's i_gid down according to an idmapping 1407 * @idmap: idmap of the mount the inode was found from 1408 * @inode: inode to map 1409 * 1410 * Return: the inode's i_gid mapped down according to @idmap. 1411 * If the inode's i_gid has no mapping INVALID_VFSGID is returned. 1412 */ 1413 static inline vfsgid_t i_gid_into_vfsgid(struct mnt_idmap *idmap, 1414 const struct inode *inode) 1415 { 1416 return make_vfsgid(idmap, i_user_ns(inode), inode->i_gid); 1417 } 1418 1419 /** 1420 * i_gid_needs_update - check whether inode's i_gid needs to be updated 1421 * @idmap: idmap of the mount the inode was found from 1422 * @attr: the new attributes of @inode 1423 * @inode: the inode to update 1424 * 1425 * Check whether the $inode's i_gid field needs to be updated taking idmapped 1426 * mounts into account if the filesystem supports it. 1427 * 1428 * Return: true if @inode's i_gid field needs to be updated, false if not. 1429 */ 1430 static inline bool i_gid_needs_update(struct mnt_idmap *idmap, 1431 const struct iattr *attr, 1432 const struct inode *inode) 1433 { 1434 return ((attr->ia_valid & ATTR_GID) && 1435 !vfsgid_eq(attr->ia_vfsgid, 1436 i_gid_into_vfsgid(idmap, inode))); 1437 } 1438 1439 /** 1440 * i_gid_update - update @inode's i_gid field 1441 * @idmap: idmap of the mount the inode was found from 1442 * @attr: the new attributes of @inode 1443 * @inode: the inode to update 1444 * 1445 * Safely update @inode's i_gid field translating the vfsgid of any idmapped 1446 * mount into the filesystem kgid. 1447 */ 1448 static inline void i_gid_update(struct mnt_idmap *idmap, 1449 const struct iattr *attr, 1450 struct inode *inode) 1451 { 1452 if (attr->ia_valid & ATTR_GID) 1453 inode->i_gid = from_vfsgid(idmap, i_user_ns(inode), 1454 attr->ia_vfsgid); 1455 } 1456 1457 /** 1458 * inode_fsuid_set - initialize inode's i_uid field with callers fsuid 1459 * @inode: inode to initialize 1460 * @idmap: idmap of the mount the inode was found from 1461 * 1462 * Initialize the i_uid field of @inode. If the inode was found/created via 1463 * an idmapped mount map the caller's fsuid according to @idmap. 1464 */ 1465 static inline void inode_fsuid_set(struct inode *inode, 1466 struct mnt_idmap *idmap) 1467 { 1468 inode->i_uid = mapped_fsuid(idmap, i_user_ns(inode)); 1469 } 1470 1471 /** 1472 * inode_fsgid_set - initialize inode's i_gid field with callers fsgid 1473 * @inode: inode to initialize 1474 * @idmap: idmap of the mount the inode was found from 1475 * 1476 * Initialize the i_gid field of @inode. If the inode was found/created via 1477 * an idmapped mount map the caller's fsgid according to @idmap. 1478 */ 1479 static inline void inode_fsgid_set(struct inode *inode, 1480 struct mnt_idmap *idmap) 1481 { 1482 inode->i_gid = mapped_fsgid(idmap, i_user_ns(inode)); 1483 } 1484 1485 /** 1486 * fsuidgid_has_mapping() - check whether caller's fsuid/fsgid is mapped 1487 * @sb: the superblock we want a mapping in 1488 * @idmap: idmap of the relevant mount 1489 * 1490 * Check whether the caller's fsuid and fsgid have a valid mapping in the 1491 * s_user_ns of the superblock @sb. If the caller is on an idmapped mount map 1492 * the caller's fsuid and fsgid according to the @idmap first. 1493 * 1494 * Return: true if fsuid and fsgid is mapped, false if not. 1495 */ 1496 static inline bool fsuidgid_has_mapping(struct super_block *sb, 1497 struct mnt_idmap *idmap) 1498 { 1499 struct user_namespace *fs_userns = sb->s_user_ns; 1500 kuid_t kuid; 1501 kgid_t kgid; 1502 1503 kuid = mapped_fsuid(idmap, fs_userns); 1504 if (!uid_valid(kuid)) 1505 return false; 1506 kgid = mapped_fsgid(idmap, fs_userns); 1507 if (!gid_valid(kgid)) 1508 return false; 1509 return kuid_has_mapping(fs_userns, kuid) && 1510 kgid_has_mapping(fs_userns, kgid); 1511 } 1512 1513 struct timespec64 current_time(struct inode *inode); 1514 struct timespec64 inode_set_ctime_current(struct inode *inode); 1515 1516 static inline time64_t inode_get_atime_sec(const struct inode *inode) 1517 { 1518 return inode->i_atime.tv_sec; 1519 } 1520 1521 static inline long inode_get_atime_nsec(const struct inode *inode) 1522 { 1523 return inode->i_atime.tv_nsec; 1524 } 1525 1526 static inline struct timespec64 inode_get_atime(const struct inode *inode) 1527 { 1528 return inode->i_atime; 1529 } 1530 1531 static inline struct timespec64 inode_set_atime_to_ts(struct inode *inode, 1532 struct timespec64 ts) 1533 { 1534 inode->i_atime = ts; 1535 return ts; 1536 } 1537 1538 static inline struct timespec64 inode_set_atime(struct inode *inode, 1539 time64_t sec, long nsec) 1540 { 1541 struct timespec64 ts = { .tv_sec = sec, 1542 .tv_nsec = nsec }; 1543 return inode_set_atime_to_ts(inode, ts); 1544 } 1545 1546 static inline time64_t inode_get_mtime_sec(const struct inode *inode) 1547 { 1548 return inode->i_mtime.tv_sec; 1549 } 1550 1551 static inline long inode_get_mtime_nsec(const struct inode *inode) 1552 { 1553 return inode->i_mtime.tv_nsec; 1554 } 1555 1556 static inline struct timespec64 inode_get_mtime(const struct inode *inode) 1557 { 1558 return inode->i_mtime; 1559 } 1560 1561 static inline struct timespec64 inode_set_mtime_to_ts(struct inode *inode, 1562 struct timespec64 ts) 1563 { 1564 inode->i_mtime = ts; 1565 return ts; 1566 } 1567 1568 static inline struct timespec64 inode_set_mtime(struct inode *inode, 1569 time64_t sec, long nsec) 1570 { 1571 struct timespec64 ts = { .tv_sec = sec, 1572 .tv_nsec = nsec }; 1573 return inode_set_mtime_to_ts(inode, ts); 1574 } 1575 1576 static inline time64_t inode_get_ctime_sec(const struct inode *inode) 1577 { 1578 return inode->__i_ctime.tv_sec; 1579 } 1580 1581 static inline long inode_get_ctime_nsec(const struct inode *inode) 1582 { 1583 return inode->__i_ctime.tv_nsec; 1584 } 1585 1586 static inline struct timespec64 inode_get_ctime(const struct inode *inode) 1587 { 1588 return inode->__i_ctime; 1589 } 1590 1591 static inline struct timespec64 inode_set_ctime_to_ts(struct inode *inode, 1592 struct timespec64 ts) 1593 { 1594 inode->__i_ctime = ts; 1595 return ts; 1596 } 1597 1598 /** 1599 * inode_set_ctime - set the ctime in the inode 1600 * @inode: inode in which to set the ctime 1601 * @sec: tv_sec value to set 1602 * @nsec: tv_nsec value to set 1603 * 1604 * Set the ctime in @inode to { @sec, @nsec } 1605 */ 1606 static inline struct timespec64 inode_set_ctime(struct inode *inode, 1607 time64_t sec, long nsec) 1608 { 1609 struct timespec64 ts = { .tv_sec = sec, 1610 .tv_nsec = nsec }; 1611 1612 return inode_set_ctime_to_ts(inode, ts); 1613 } 1614 1615 struct timespec64 simple_inode_init_ts(struct inode *inode); 1616 1617 /* 1618 * Snapshotting support. 1619 */ 1620 1621 /* 1622 * These are internal functions, please use sb_start_{write,pagefault,intwrite} 1623 * instead. 1624 */ 1625 static inline void __sb_end_write(struct super_block *sb, int level) 1626 { 1627 percpu_up_read(sb->s_writers.rw_sem + level-1); 1628 } 1629 1630 static inline void __sb_start_write(struct super_block *sb, int level) 1631 { 1632 percpu_down_read(sb->s_writers.rw_sem + level - 1); 1633 } 1634 1635 static inline bool __sb_start_write_trylock(struct super_block *sb, int level) 1636 { 1637 return percpu_down_read_trylock(sb->s_writers.rw_sem + level - 1); 1638 } 1639 1640 #define __sb_writers_acquired(sb, lev) \ 1641 percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_) 1642 #define __sb_writers_release(sb, lev) \ 1643 percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_) 1644 1645 static inline bool sb_write_started(const struct super_block *sb) 1646 { 1647 return lockdep_is_held_type(sb->s_writers.rw_sem + SB_FREEZE_WRITE - 1, 1); 1648 } 1649 1650 /** 1651 * sb_end_write - drop write access to a superblock 1652 * @sb: the super we wrote to 1653 * 1654 * Decrement number of writers to the filesystem. Wake up possible waiters 1655 * wanting to freeze the filesystem. 1656 */ 1657 static inline void sb_end_write(struct super_block *sb) 1658 { 1659 __sb_end_write(sb, SB_FREEZE_WRITE); 1660 } 1661 1662 /** 1663 * sb_end_pagefault - drop write access to a superblock from a page fault 1664 * @sb: the super we wrote to 1665 * 1666 * Decrement number of processes handling write page fault to the filesystem. 1667 * Wake up possible waiters wanting to freeze the filesystem. 1668 */ 1669 static inline void sb_end_pagefault(struct super_block *sb) 1670 { 1671 __sb_end_write(sb, SB_FREEZE_PAGEFAULT); 1672 } 1673 1674 /** 1675 * sb_end_intwrite - drop write access to a superblock for internal fs purposes 1676 * @sb: the super we wrote to 1677 * 1678 * Decrement fs-internal number of writers to the filesystem. Wake up possible 1679 * waiters wanting to freeze the filesystem. 1680 */ 1681 static inline void sb_end_intwrite(struct super_block *sb) 1682 { 1683 __sb_end_write(sb, SB_FREEZE_FS); 1684 } 1685 1686 /** 1687 * sb_start_write - get write access to a superblock 1688 * @sb: the super we write to 1689 * 1690 * When a process wants to write data or metadata to a file system (i.e. dirty 1691 * a page or an inode), it should embed the operation in a sb_start_write() - 1692 * sb_end_write() pair to get exclusion against file system freezing. This 1693 * function increments number of writers preventing freezing. If the file 1694 * system is already frozen, the function waits until the file system is 1695 * thawed. 1696 * 1697 * Since freeze protection behaves as a lock, users have to preserve 1698 * ordering of freeze protection and other filesystem locks. Generally, 1699 * freeze protection should be the outermost lock. In particular, we have: 1700 * 1701 * sb_start_write 1702 * -> i_mutex (write path, truncate, directory ops, ...) 1703 * -> s_umount (freeze_super, thaw_super) 1704 */ 1705 static inline void sb_start_write(struct super_block *sb) 1706 { 1707 __sb_start_write(sb, SB_FREEZE_WRITE); 1708 } 1709 1710 static inline bool sb_start_write_trylock(struct super_block *sb) 1711 { 1712 return __sb_start_write_trylock(sb, SB_FREEZE_WRITE); 1713 } 1714 1715 /** 1716 * sb_start_pagefault - get write access to a superblock from a page fault 1717 * @sb: the super we write to 1718 * 1719 * When a process starts handling write page fault, it should embed the 1720 * operation into sb_start_pagefault() - sb_end_pagefault() pair to get 1721 * exclusion against file system freezing. This is needed since the page fault 1722 * is going to dirty a page. This function increments number of running page 1723 * faults preventing freezing. If the file system is already frozen, the 1724 * function waits until the file system is thawed. 1725 * 1726 * Since page fault freeze protection behaves as a lock, users have to preserve 1727 * ordering of freeze protection and other filesystem locks. It is advised to 1728 * put sb_start_pagefault() close to mmap_lock in lock ordering. Page fault 1729 * handling code implies lock dependency: 1730 * 1731 * mmap_lock 1732 * -> sb_start_pagefault 1733 */ 1734 static inline void sb_start_pagefault(struct super_block *sb) 1735 { 1736 __sb_start_write(sb, SB_FREEZE_PAGEFAULT); 1737 } 1738 1739 /** 1740 * sb_start_intwrite - get write access to a superblock for internal fs purposes 1741 * @sb: the super we write to 1742 * 1743 * This is the third level of protection against filesystem freezing. It is 1744 * free for use by a filesystem. The only requirement is that it must rank 1745 * below sb_start_pagefault. 1746 * 1747 * For example filesystem can call sb_start_intwrite() when starting a 1748 * transaction which somewhat eases handling of freezing for internal sources 1749 * of filesystem changes (internal fs threads, discarding preallocation on file 1750 * close, etc.). 1751 */ 1752 static inline void sb_start_intwrite(struct super_block *sb) 1753 { 1754 __sb_start_write(sb, SB_FREEZE_FS); 1755 } 1756 1757 static inline bool sb_start_intwrite_trylock(struct super_block *sb) 1758 { 1759 return __sb_start_write_trylock(sb, SB_FREEZE_FS); 1760 } 1761 1762 bool inode_owner_or_capable(struct mnt_idmap *idmap, 1763 const struct inode *inode); 1764 1765 /* 1766 * VFS helper functions.. 1767 */ 1768 int vfs_create(struct mnt_idmap *, struct inode *, 1769 struct dentry *, umode_t, bool); 1770 int vfs_mkdir(struct mnt_idmap *, struct inode *, 1771 struct dentry *, umode_t); 1772 int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *, 1773 umode_t, dev_t); 1774 int vfs_symlink(struct mnt_idmap *, struct inode *, 1775 struct dentry *, const char *); 1776 int vfs_link(struct dentry *, struct mnt_idmap *, struct inode *, 1777 struct dentry *, struct inode **); 1778 int vfs_rmdir(struct mnt_idmap *, struct inode *, struct dentry *); 1779 int vfs_unlink(struct mnt_idmap *, struct inode *, struct dentry *, 1780 struct inode **); 1781 1782 /** 1783 * struct renamedata - contains all information required for renaming 1784 * @old_mnt_idmap: idmap of the old mount the inode was found from 1785 * @old_dir: parent of source 1786 * @old_dentry: source 1787 * @new_mnt_idmap: idmap of the new mount the inode was found from 1788 * @new_dir: parent of destination 1789 * @new_dentry: destination 1790 * @delegated_inode: returns an inode needing a delegation break 1791 * @flags: rename flags 1792 */ 1793 struct renamedata { 1794 struct mnt_idmap *old_mnt_idmap; 1795 struct inode *old_dir; 1796 struct dentry *old_dentry; 1797 struct mnt_idmap *new_mnt_idmap; 1798 struct inode *new_dir; 1799 struct dentry *new_dentry; 1800 struct inode **delegated_inode; 1801 unsigned int flags; 1802 } __randomize_layout; 1803 1804 int vfs_rename(struct renamedata *); 1805 1806 static inline int vfs_whiteout(struct mnt_idmap *idmap, 1807 struct inode *dir, struct dentry *dentry) 1808 { 1809 return vfs_mknod(idmap, dir, dentry, S_IFCHR | WHITEOUT_MODE, 1810 WHITEOUT_DEV); 1811 } 1812 1813 struct file *kernel_tmpfile_open(struct mnt_idmap *idmap, 1814 const struct path *parentpath, 1815 umode_t mode, int open_flag, 1816 const struct cred *cred); 1817 struct file *kernel_file_open(const struct path *path, int flags, 1818 struct inode *inode, const struct cred *cred); 1819 1820 int vfs_mkobj(struct dentry *, umode_t, 1821 int (*f)(struct dentry *, umode_t, void *), 1822 void *); 1823 1824 int vfs_fchown(struct file *file, uid_t user, gid_t group); 1825 int vfs_fchmod(struct file *file, umode_t mode); 1826 int vfs_utimes(const struct path *path, struct timespec64 *times); 1827 1828 extern long vfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 1829 1830 #ifdef CONFIG_COMPAT 1831 extern long compat_ptr_ioctl(struct file *file, unsigned int cmd, 1832 unsigned long arg); 1833 #else 1834 #define compat_ptr_ioctl NULL 1835 #endif 1836 1837 /* 1838 * VFS file helper functions. 1839 */ 1840 void inode_init_owner(struct mnt_idmap *idmap, struct inode *inode, 1841 const struct inode *dir, umode_t mode); 1842 extern bool may_open_dev(const struct path *path); 1843 umode_t mode_strip_sgid(struct mnt_idmap *idmap, 1844 const struct inode *dir, umode_t mode); 1845 1846 /* 1847 * This is the "filldir" function type, used by readdir() to let 1848 * the kernel specify what kind of dirent layout it wants to have. 1849 * This allows the kernel to read directories into kernel space or 1850 * to have different dirent layouts depending on the binary type. 1851 * Return 'true' to keep going and 'false' if there are no more entries. 1852 */ 1853 struct dir_context; 1854 typedef bool (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64, 1855 unsigned); 1856 1857 struct dir_context { 1858 filldir_t actor; 1859 loff_t pos; 1860 }; 1861 1862 /* 1863 * These flags let !MMU mmap() govern direct device mapping vs immediate 1864 * copying more easily for MAP_PRIVATE, especially for ROM filesystems. 1865 * 1866 * NOMMU_MAP_COPY: Copy can be mapped (MAP_PRIVATE) 1867 * NOMMU_MAP_DIRECT: Can be mapped directly (MAP_SHARED) 1868 * NOMMU_MAP_READ: Can be mapped for reading 1869 * NOMMU_MAP_WRITE: Can be mapped for writing 1870 * NOMMU_MAP_EXEC: Can be mapped for execution 1871 */ 1872 #define NOMMU_MAP_COPY 0x00000001 1873 #define NOMMU_MAP_DIRECT 0x00000008 1874 #define NOMMU_MAP_READ VM_MAYREAD 1875 #define NOMMU_MAP_WRITE VM_MAYWRITE 1876 #define NOMMU_MAP_EXEC VM_MAYEXEC 1877 1878 #define NOMMU_VMFLAGS \ 1879 (NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC) 1880 1881 /* 1882 * These flags control the behavior of the remap_file_range function pointer. 1883 * If it is called with len == 0 that means "remap to end of source file". 1884 * See Documentation/filesystems/vfs.rst for more details about this call. 1885 * 1886 * REMAP_FILE_DEDUP: only remap if contents identical (i.e. deduplicate) 1887 * REMAP_FILE_CAN_SHORTEN: caller can handle a shortened request 1888 */ 1889 #define REMAP_FILE_DEDUP (1 << 0) 1890 #define REMAP_FILE_CAN_SHORTEN (1 << 1) 1891 1892 /* 1893 * These flags signal that the caller is ok with altering various aspects of 1894 * the behavior of the remap operation. The changes must be made by the 1895 * implementation; the vfs remap helper functions can take advantage of them. 1896 * Flags in this category exist to preserve the quirky behavior of the hoisted 1897 * btrfs clone/dedupe ioctls. 1898 */ 1899 #define REMAP_FILE_ADVISORY (REMAP_FILE_CAN_SHORTEN) 1900 1901 /* 1902 * These flags control the behavior of vfs_copy_file_range(). 1903 * They are not available to the user via syscall. 1904 * 1905 * COPY_FILE_SPLICE: call splice direct instead of fs clone/copy ops 1906 */ 1907 #define COPY_FILE_SPLICE (1 << 0) 1908 1909 struct iov_iter; 1910 struct io_uring_cmd; 1911 struct offset_ctx; 1912 1913 struct file_operations { 1914 struct module *owner; 1915 loff_t (*llseek) (struct file *, loff_t, int); 1916 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *); 1917 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *); 1918 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *); 1919 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *); 1920 int (*iopoll)(struct kiocb *kiocb, struct io_comp_batch *, 1921 unsigned int flags); 1922 int (*iterate_shared) (struct file *, struct dir_context *); 1923 __poll_t (*poll) (struct file *, struct poll_table_struct *); 1924 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long); 1925 long (*compat_ioctl) (struct file *, unsigned int, unsigned long); 1926 int (*mmap) (struct file *, struct vm_area_struct *); 1927 unsigned long mmap_supported_flags; 1928 int (*open) (struct inode *, struct file *); 1929 int (*flush) (struct file *, fl_owner_t id); 1930 int (*release) (struct inode *, struct file *); 1931 int (*fsync) (struct file *, loff_t, loff_t, int datasync); 1932 int (*fasync) (int, struct file *, int); 1933 int (*lock) (struct file *, int, struct file_lock *); 1934 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 1935 int (*check_flags)(int); 1936 int (*flock) (struct file *, int, struct file_lock *); 1937 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int); 1938 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int); 1939 void (*splice_eof)(struct file *file); 1940 int (*setlease)(struct file *, int, struct file_lock **, void **); 1941 long (*fallocate)(struct file *file, int mode, loff_t offset, 1942 loff_t len); 1943 void (*show_fdinfo)(struct seq_file *m, struct file *f); 1944 #ifndef CONFIG_MMU 1945 unsigned (*mmap_capabilities)(struct file *); 1946 #endif 1947 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *, 1948 loff_t, size_t, unsigned int); 1949 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in, 1950 struct file *file_out, loff_t pos_out, 1951 loff_t len, unsigned int remap_flags); 1952 int (*fadvise)(struct file *, loff_t, loff_t, int); 1953 int (*uring_cmd)(struct io_uring_cmd *ioucmd, unsigned int issue_flags); 1954 int (*uring_cmd_iopoll)(struct io_uring_cmd *, struct io_comp_batch *, 1955 unsigned int poll_flags); 1956 } __randomize_layout; 1957 1958 /* Wrap a directory iterator that needs exclusive inode access */ 1959 int wrap_directory_iterator(struct file *, struct dir_context *, 1960 int (*) (struct file *, struct dir_context *)); 1961 #define WRAP_DIR_ITER(x) \ 1962 static int shared_##x(struct file *file , struct dir_context *ctx) \ 1963 { return wrap_directory_iterator(file, ctx, x); } 1964 1965 struct inode_operations { 1966 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int); 1967 const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *); 1968 int (*permission) (struct mnt_idmap *, struct inode *, int); 1969 struct posix_acl * (*get_inode_acl)(struct inode *, int, bool); 1970 1971 int (*readlink) (struct dentry *, char __user *,int); 1972 1973 int (*create) (struct mnt_idmap *, struct inode *,struct dentry *, 1974 umode_t, bool); 1975 int (*link) (struct dentry *,struct inode *,struct dentry *); 1976 int (*unlink) (struct inode *,struct dentry *); 1977 int (*symlink) (struct mnt_idmap *, struct inode *,struct dentry *, 1978 const char *); 1979 int (*mkdir) (struct mnt_idmap *, struct inode *,struct dentry *, 1980 umode_t); 1981 int (*rmdir) (struct inode *,struct dentry *); 1982 int (*mknod) (struct mnt_idmap *, struct inode *,struct dentry *, 1983 umode_t,dev_t); 1984 int (*rename) (struct mnt_idmap *, struct inode *, struct dentry *, 1985 struct inode *, struct dentry *, unsigned int); 1986 int (*setattr) (struct mnt_idmap *, struct dentry *, struct iattr *); 1987 int (*getattr) (struct mnt_idmap *, const struct path *, 1988 struct kstat *, u32, unsigned int); 1989 ssize_t (*listxattr) (struct dentry *, char *, size_t); 1990 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start, 1991 u64 len); 1992 int (*update_time)(struct inode *, int); 1993 int (*atomic_open)(struct inode *, struct dentry *, 1994 struct file *, unsigned open_flag, 1995 umode_t create_mode); 1996 int (*tmpfile) (struct mnt_idmap *, struct inode *, 1997 struct file *, umode_t); 1998 struct posix_acl *(*get_acl)(struct mnt_idmap *, struct dentry *, 1999 int); 2000 int (*set_acl)(struct mnt_idmap *, struct dentry *, 2001 struct posix_acl *, int); 2002 int (*fileattr_set)(struct mnt_idmap *idmap, 2003 struct dentry *dentry, struct fileattr *fa); 2004 int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa); 2005 struct offset_ctx *(*get_offset_ctx)(struct inode *inode); 2006 } ____cacheline_aligned; 2007 2008 static inline ssize_t call_read_iter(struct file *file, struct kiocb *kio, 2009 struct iov_iter *iter) 2010 { 2011 return file->f_op->read_iter(kio, iter); 2012 } 2013 2014 static inline ssize_t call_write_iter(struct file *file, struct kiocb *kio, 2015 struct iov_iter *iter) 2016 { 2017 return file->f_op->write_iter(kio, iter); 2018 } 2019 2020 static inline int call_mmap(struct file *file, struct vm_area_struct *vma) 2021 { 2022 return file->f_op->mmap(file, vma); 2023 } 2024 2025 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *); 2026 extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *); 2027 extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *, 2028 loff_t, size_t, unsigned int); 2029 extern ssize_t generic_copy_file_range(struct file *file_in, loff_t pos_in, 2030 struct file *file_out, loff_t pos_out, 2031 size_t len, unsigned int flags); 2032 int __generic_remap_file_range_prep(struct file *file_in, loff_t pos_in, 2033 struct file *file_out, loff_t pos_out, 2034 loff_t *len, unsigned int remap_flags, 2035 const struct iomap_ops *dax_read_ops); 2036 int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in, 2037 struct file *file_out, loff_t pos_out, 2038 loff_t *count, unsigned int remap_flags); 2039 extern loff_t do_clone_file_range(struct file *file_in, loff_t pos_in, 2040 struct file *file_out, loff_t pos_out, 2041 loff_t len, unsigned int remap_flags); 2042 extern loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in, 2043 struct file *file_out, loff_t pos_out, 2044 loff_t len, unsigned int remap_flags); 2045 extern int vfs_dedupe_file_range(struct file *file, 2046 struct file_dedupe_range *same); 2047 extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos, 2048 struct file *dst_file, loff_t dst_pos, 2049 loff_t len, unsigned int remap_flags); 2050 2051 enum freeze_holder { 2052 FREEZE_HOLDER_KERNEL = (1U << 0), 2053 FREEZE_HOLDER_USERSPACE = (1U << 1), 2054 }; 2055 2056 struct super_operations { 2057 struct inode *(*alloc_inode)(struct super_block *sb); 2058 void (*destroy_inode)(struct inode *); 2059 void (*free_inode)(struct inode *); 2060 2061 void (*dirty_inode) (struct inode *, int flags); 2062 int (*write_inode) (struct inode *, struct writeback_control *wbc); 2063 int (*drop_inode) (struct inode *); 2064 void (*evict_inode) (struct inode *); 2065 void (*put_super) (struct super_block *); 2066 int (*sync_fs)(struct super_block *sb, int wait); 2067 int (*freeze_super) (struct super_block *, enum freeze_holder who); 2068 int (*freeze_fs) (struct super_block *); 2069 int (*thaw_super) (struct super_block *, enum freeze_holder who); 2070 int (*unfreeze_fs) (struct super_block *); 2071 int (*statfs) (struct dentry *, struct kstatfs *); 2072 int (*remount_fs) (struct super_block *, int *, char *); 2073 void (*umount_begin) (struct super_block *); 2074 2075 int (*show_options)(struct seq_file *, struct dentry *); 2076 int (*show_devname)(struct seq_file *, struct dentry *); 2077 int (*show_path)(struct seq_file *, struct dentry *); 2078 int (*show_stats)(struct seq_file *, struct dentry *); 2079 #ifdef CONFIG_QUOTA 2080 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t); 2081 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t); 2082 struct dquot __rcu **(*get_dquots)(struct inode *); 2083 #endif 2084 long (*nr_cached_objects)(struct super_block *, 2085 struct shrink_control *); 2086 long (*free_cached_objects)(struct super_block *, 2087 struct shrink_control *); 2088 void (*shutdown)(struct super_block *sb); 2089 }; 2090 2091 /* 2092 * Inode flags - they have no relation to superblock flags now 2093 */ 2094 #define S_SYNC (1 << 0) /* Writes are synced at once */ 2095 #define S_NOATIME (1 << 1) /* Do not update access times */ 2096 #define S_APPEND (1 << 2) /* Append-only file */ 2097 #define S_IMMUTABLE (1 << 3) /* Immutable file */ 2098 #define S_DEAD (1 << 4) /* removed, but still open directory */ 2099 #define S_NOQUOTA (1 << 5) /* Inode is not counted to quota */ 2100 #define S_DIRSYNC (1 << 6) /* Directory modifications are synchronous */ 2101 #define S_NOCMTIME (1 << 7) /* Do not update file c/mtime */ 2102 #define S_SWAPFILE (1 << 8) /* Do not truncate: swapon got its bmaps */ 2103 #define S_PRIVATE (1 << 9) /* Inode is fs-internal */ 2104 #define S_IMA (1 << 10) /* Inode has an associated IMA struct */ 2105 #define S_AUTOMOUNT (1 << 11) /* Automount/referral quasi-directory */ 2106 #define S_NOSEC (1 << 12) /* no suid or xattr security attributes */ 2107 #ifdef CONFIG_FS_DAX 2108 #define S_DAX (1 << 13) /* Direct Access, avoiding the page cache */ 2109 #else 2110 #define S_DAX 0 /* Make all the DAX code disappear */ 2111 #endif 2112 #define S_ENCRYPTED (1 << 14) /* Encrypted file (using fs/crypto/) */ 2113 #define S_CASEFOLD (1 << 15) /* Casefolded file */ 2114 #define S_VERITY (1 << 16) /* Verity file (using fs/verity/) */ 2115 #define S_KERNEL_FILE (1 << 17) /* File is in use by the kernel (eg. fs/cachefiles) */ 2116 2117 /* 2118 * Note that nosuid etc flags are inode-specific: setting some file-system 2119 * flags just means all the inodes inherit those flags by default. It might be 2120 * possible to override it selectively if you really wanted to with some 2121 * ioctl() that is not currently implemented. 2122 * 2123 * Exception: SB_RDONLY is always applied to the entire file system. 2124 * 2125 * Unfortunately, it is possible to change a filesystems flags with it mounted 2126 * with files in use. This means that all of the inodes will not have their 2127 * i_flags updated. Hence, i_flags no longer inherit the superblock mount 2128 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org 2129 */ 2130 #define __IS_FLG(inode, flg) ((inode)->i_sb->s_flags & (flg)) 2131 2132 static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; } 2133 #define IS_RDONLY(inode) sb_rdonly((inode)->i_sb) 2134 #define IS_SYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS) || \ 2135 ((inode)->i_flags & S_SYNC)) 2136 #define IS_DIRSYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS|SB_DIRSYNC) || \ 2137 ((inode)->i_flags & (S_SYNC|S_DIRSYNC))) 2138 #define IS_MANDLOCK(inode) __IS_FLG(inode, SB_MANDLOCK) 2139 #define IS_NOATIME(inode) __IS_FLG(inode, SB_RDONLY|SB_NOATIME) 2140 #define IS_I_VERSION(inode) __IS_FLG(inode, SB_I_VERSION) 2141 2142 #define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA) 2143 #define IS_APPEND(inode) ((inode)->i_flags & S_APPEND) 2144 #define IS_IMMUTABLE(inode) ((inode)->i_flags & S_IMMUTABLE) 2145 #define IS_POSIXACL(inode) __IS_FLG(inode, SB_POSIXACL) 2146 2147 #define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD) 2148 #define IS_NOCMTIME(inode) ((inode)->i_flags & S_NOCMTIME) 2149 #define IS_SWAPFILE(inode) ((inode)->i_flags & S_SWAPFILE) 2150 #define IS_PRIVATE(inode) ((inode)->i_flags & S_PRIVATE) 2151 #define IS_IMA(inode) ((inode)->i_flags & S_IMA) 2152 #define IS_AUTOMOUNT(inode) ((inode)->i_flags & S_AUTOMOUNT) 2153 #define IS_NOSEC(inode) ((inode)->i_flags & S_NOSEC) 2154 #define IS_DAX(inode) ((inode)->i_flags & S_DAX) 2155 #define IS_ENCRYPTED(inode) ((inode)->i_flags & S_ENCRYPTED) 2156 #define IS_CASEFOLDED(inode) ((inode)->i_flags & S_CASEFOLD) 2157 #define IS_VERITY(inode) ((inode)->i_flags & S_VERITY) 2158 2159 #define IS_WHITEOUT(inode) (S_ISCHR(inode->i_mode) && \ 2160 (inode)->i_rdev == WHITEOUT_DEV) 2161 2162 static inline bool HAS_UNMAPPED_ID(struct mnt_idmap *idmap, 2163 struct inode *inode) 2164 { 2165 return !vfsuid_valid(i_uid_into_vfsuid(idmap, inode)) || 2166 !vfsgid_valid(i_gid_into_vfsgid(idmap, inode)); 2167 } 2168 2169 static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp) 2170 { 2171 *kiocb = (struct kiocb) { 2172 .ki_filp = filp, 2173 .ki_flags = filp->f_iocb_flags, 2174 .ki_ioprio = get_current_ioprio(), 2175 }; 2176 } 2177 2178 static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src, 2179 struct file *filp) 2180 { 2181 *kiocb = (struct kiocb) { 2182 .ki_filp = filp, 2183 .ki_flags = kiocb_src->ki_flags, 2184 .ki_ioprio = kiocb_src->ki_ioprio, 2185 .ki_pos = kiocb_src->ki_pos, 2186 }; 2187 } 2188 2189 /* 2190 * Inode state bits. Protected by inode->i_lock 2191 * 2192 * Four bits determine the dirty state of the inode: I_DIRTY_SYNC, 2193 * I_DIRTY_DATASYNC, I_DIRTY_PAGES, and I_DIRTY_TIME. 2194 * 2195 * Four bits define the lifetime of an inode. Initially, inodes are I_NEW, 2196 * until that flag is cleared. I_WILL_FREE, I_FREEING and I_CLEAR are set at 2197 * various stages of removing an inode. 2198 * 2199 * Two bits are used for locking and completion notification, I_NEW and I_SYNC. 2200 * 2201 * I_DIRTY_SYNC Inode is dirty, but doesn't have to be written on 2202 * fdatasync() (unless I_DIRTY_DATASYNC is also set). 2203 * Timestamp updates are the usual cause. 2204 * I_DIRTY_DATASYNC Data-related inode changes pending. We keep track of 2205 * these changes separately from I_DIRTY_SYNC so that we 2206 * don't have to write inode on fdatasync() when only 2207 * e.g. the timestamps have changed. 2208 * I_DIRTY_PAGES Inode has dirty pages. Inode itself may be clean. 2209 * I_DIRTY_TIME The inode itself has dirty timestamps, and the 2210 * lazytime mount option is enabled. We keep track of this 2211 * separately from I_DIRTY_SYNC in order to implement 2212 * lazytime. This gets cleared if I_DIRTY_INODE 2213 * (I_DIRTY_SYNC and/or I_DIRTY_DATASYNC) gets set. But 2214 * I_DIRTY_TIME can still be set if I_DIRTY_SYNC is already 2215 * in place because writeback might already be in progress 2216 * and we don't want to lose the time update 2217 * I_NEW Serves as both a mutex and completion notification. 2218 * New inodes set I_NEW. If two processes both create 2219 * the same inode, one of them will release its inode and 2220 * wait for I_NEW to be released before returning. 2221 * Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can 2222 * also cause waiting on I_NEW, without I_NEW actually 2223 * being set. find_inode() uses this to prevent returning 2224 * nearly-dead inodes. 2225 * I_WILL_FREE Must be set when calling write_inode_now() if i_count 2226 * is zero. I_FREEING must be set when I_WILL_FREE is 2227 * cleared. 2228 * I_FREEING Set when inode is about to be freed but still has dirty 2229 * pages or buffers attached or the inode itself is still 2230 * dirty. 2231 * I_CLEAR Added by clear_inode(). In this state the inode is 2232 * clean and can be destroyed. Inode keeps I_FREEING. 2233 * 2234 * Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are 2235 * prohibited for many purposes. iget() must wait for 2236 * the inode to be completely released, then create it 2237 * anew. Other functions will just ignore such inodes, 2238 * if appropriate. I_NEW is used for waiting. 2239 * 2240 * I_SYNC Writeback of inode is running. The bit is set during 2241 * data writeback, and cleared with a wakeup on the bit 2242 * address once it is done. The bit is also used to pin 2243 * the inode in memory for flusher thread. 2244 * 2245 * I_REFERENCED Marks the inode as recently references on the LRU list. 2246 * 2247 * I_DIO_WAKEUP Never set. Only used as a key for wait_on_bit(). 2248 * 2249 * I_WB_SWITCH Cgroup bdi_writeback switching in progress. Used to 2250 * synchronize competing switching instances and to tell 2251 * wb stat updates to grab the i_pages lock. See 2252 * inode_switch_wbs_work_fn() for details. 2253 * 2254 * I_OVL_INUSE Used by overlayfs to get exclusive ownership on upper 2255 * and work dirs among overlayfs mounts. 2256 * 2257 * I_CREATING New object's inode in the middle of setting up. 2258 * 2259 * I_DONTCACHE Evict inode as soon as it is not used anymore. 2260 * 2261 * I_SYNC_QUEUED Inode is queued in b_io or b_more_io writeback lists. 2262 * Used to detect that mark_inode_dirty() should not move 2263 * inode between dirty lists. 2264 * 2265 * I_PINNING_FSCACHE_WB Inode is pinning an fscache object for writeback. 2266 * 2267 * Q: What is the difference between I_WILL_FREE and I_FREEING? 2268 */ 2269 #define I_DIRTY_SYNC (1 << 0) 2270 #define I_DIRTY_DATASYNC (1 << 1) 2271 #define I_DIRTY_PAGES (1 << 2) 2272 #define __I_NEW 3 2273 #define I_NEW (1 << __I_NEW) 2274 #define I_WILL_FREE (1 << 4) 2275 #define I_FREEING (1 << 5) 2276 #define I_CLEAR (1 << 6) 2277 #define __I_SYNC 7 2278 #define I_SYNC (1 << __I_SYNC) 2279 #define I_REFERENCED (1 << 8) 2280 #define __I_DIO_WAKEUP 9 2281 #define I_DIO_WAKEUP (1 << __I_DIO_WAKEUP) 2282 #define I_LINKABLE (1 << 10) 2283 #define I_DIRTY_TIME (1 << 11) 2284 #define I_WB_SWITCH (1 << 13) 2285 #define I_OVL_INUSE (1 << 14) 2286 #define I_CREATING (1 << 15) 2287 #define I_DONTCACHE (1 << 16) 2288 #define I_SYNC_QUEUED (1 << 17) 2289 #define I_PINNING_FSCACHE_WB (1 << 18) 2290 2291 #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC) 2292 #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES) 2293 #define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME) 2294 2295 extern void __mark_inode_dirty(struct inode *, int); 2296 static inline void mark_inode_dirty(struct inode *inode) 2297 { 2298 __mark_inode_dirty(inode, I_DIRTY); 2299 } 2300 2301 static inline void mark_inode_dirty_sync(struct inode *inode) 2302 { 2303 __mark_inode_dirty(inode, I_DIRTY_SYNC); 2304 } 2305 2306 /* 2307 * Returns true if the given inode itself only has dirty timestamps (its pages 2308 * may still be dirty) and isn't currently being allocated or freed. 2309 * Filesystems should call this if when writing an inode when lazytime is 2310 * enabled, they want to opportunistically write the timestamps of other inodes 2311 * located very nearby on-disk, e.g. in the same inode block. This returns true 2312 * if the given inode is in need of such an opportunistic update. Requires 2313 * i_lock, or at least later re-checking under i_lock. 2314 */ 2315 static inline bool inode_is_dirtytime_only(struct inode *inode) 2316 { 2317 return (inode->i_state & (I_DIRTY_TIME | I_NEW | 2318 I_FREEING | I_WILL_FREE)) == I_DIRTY_TIME; 2319 } 2320 2321 extern void inc_nlink(struct inode *inode); 2322 extern void drop_nlink(struct inode *inode); 2323 extern void clear_nlink(struct inode *inode); 2324 extern void set_nlink(struct inode *inode, unsigned int nlink); 2325 2326 static inline void inode_inc_link_count(struct inode *inode) 2327 { 2328 inc_nlink(inode); 2329 mark_inode_dirty(inode); 2330 } 2331 2332 static inline void inode_dec_link_count(struct inode *inode) 2333 { 2334 drop_nlink(inode); 2335 mark_inode_dirty(inode); 2336 } 2337 2338 enum file_time_flags { 2339 S_ATIME = 1, 2340 S_MTIME = 2, 2341 S_CTIME = 4, 2342 S_VERSION = 8, 2343 }; 2344 2345 extern bool atime_needs_update(const struct path *, struct inode *); 2346 extern void touch_atime(const struct path *); 2347 int inode_update_time(struct inode *inode, int flags); 2348 2349 static inline void file_accessed(struct file *file) 2350 { 2351 if (!(file->f_flags & O_NOATIME)) 2352 touch_atime(&file->f_path); 2353 } 2354 2355 extern int file_modified(struct file *file); 2356 int kiocb_modified(struct kiocb *iocb); 2357 2358 int sync_inode_metadata(struct inode *inode, int wait); 2359 2360 struct file_system_type { 2361 const char *name; 2362 int fs_flags; 2363 #define FS_REQUIRES_DEV 1 2364 #define FS_BINARY_MOUNTDATA 2 2365 #define FS_HAS_SUBTYPE 4 2366 #define FS_USERNS_MOUNT 8 /* Can be mounted by userns root */ 2367 #define FS_DISALLOW_NOTIFY_PERM 16 /* Disable fanotify permission events */ 2368 #define FS_ALLOW_IDMAP 32 /* FS has been updated to handle vfs idmappings. */ 2369 #define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() during rename() internally. */ 2370 int (*init_fs_context)(struct fs_context *); 2371 const struct fs_parameter_spec *parameters; 2372 struct dentry *(*mount) (struct file_system_type *, int, 2373 const char *, void *); 2374 void (*kill_sb) (struct super_block *); 2375 struct module *owner; 2376 struct file_system_type * next; 2377 struct hlist_head fs_supers; 2378 2379 struct lock_class_key s_lock_key; 2380 struct lock_class_key s_umount_key; 2381 struct lock_class_key s_vfs_rename_key; 2382 struct lock_class_key s_writers_key[SB_FREEZE_LEVELS]; 2383 2384 struct lock_class_key i_lock_key; 2385 struct lock_class_key i_mutex_key; 2386 struct lock_class_key invalidate_lock_key; 2387 struct lock_class_key i_mutex_dir_key; 2388 }; 2389 2390 #define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME) 2391 2392 extern struct dentry *mount_bdev(struct file_system_type *fs_type, 2393 int flags, const char *dev_name, void *data, 2394 int (*fill_super)(struct super_block *, void *, int)); 2395 extern struct dentry *mount_single(struct file_system_type *fs_type, 2396 int flags, void *data, 2397 int (*fill_super)(struct super_block *, void *, int)); 2398 extern struct dentry *mount_nodev(struct file_system_type *fs_type, 2399 int flags, void *data, 2400 int (*fill_super)(struct super_block *, void *, int)); 2401 extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path); 2402 void retire_super(struct super_block *sb); 2403 void generic_shutdown_super(struct super_block *sb); 2404 void kill_block_super(struct super_block *sb); 2405 void kill_anon_super(struct super_block *sb); 2406 void kill_litter_super(struct super_block *sb); 2407 void deactivate_super(struct super_block *sb); 2408 void deactivate_locked_super(struct super_block *sb); 2409 int set_anon_super(struct super_block *s, void *data); 2410 int set_anon_super_fc(struct super_block *s, struct fs_context *fc); 2411 int get_anon_bdev(dev_t *); 2412 void free_anon_bdev(dev_t); 2413 struct super_block *sget_fc(struct fs_context *fc, 2414 int (*test)(struct super_block *, struct fs_context *), 2415 int (*set)(struct super_block *, struct fs_context *)); 2416 struct super_block *sget(struct file_system_type *type, 2417 int (*test)(struct super_block *,void *), 2418 int (*set)(struct super_block *,void *), 2419 int flags, void *data); 2420 struct super_block *sget_dev(struct fs_context *fc, dev_t dev); 2421 2422 /* Alas, no aliases. Too much hassle with bringing module.h everywhere */ 2423 #define fops_get(fops) \ 2424 (((fops) && try_module_get((fops)->owner) ? (fops) : NULL)) 2425 #define fops_put(fops) \ 2426 do { if (fops) module_put((fops)->owner); } while(0) 2427 /* 2428 * This one is to be used *ONLY* from ->open() instances. 2429 * fops must be non-NULL, pinned down *and* module dependencies 2430 * should be sufficient to pin the caller down as well. 2431 */ 2432 #define replace_fops(f, fops) \ 2433 do { \ 2434 struct file *__file = (f); \ 2435 fops_put(__file->f_op); \ 2436 BUG_ON(!(__file->f_op = (fops))); \ 2437 } while(0) 2438 2439 extern int register_filesystem(struct file_system_type *); 2440 extern int unregister_filesystem(struct file_system_type *); 2441 extern int vfs_statfs(const struct path *, struct kstatfs *); 2442 extern int user_statfs(const char __user *, struct kstatfs *); 2443 extern int fd_statfs(int, struct kstatfs *); 2444 int freeze_super(struct super_block *super, enum freeze_holder who); 2445 int thaw_super(struct super_block *super, enum freeze_holder who); 2446 extern __printf(2, 3) 2447 int super_setup_bdi_name(struct super_block *sb, char *fmt, ...); 2448 extern int super_setup_bdi(struct super_block *sb); 2449 2450 extern int current_umask(void); 2451 2452 extern void ihold(struct inode * inode); 2453 extern void iput(struct inode *); 2454 int inode_update_timestamps(struct inode *inode, int flags); 2455 int generic_update_time(struct inode *, int); 2456 2457 /* /sys/fs */ 2458 extern struct kobject *fs_kobj; 2459 2460 #define MAX_RW_COUNT (INT_MAX & PAGE_MASK) 2461 2462 /* fs/open.c */ 2463 struct audit_names; 2464 struct filename { 2465 const char *name; /* pointer to actual string */ 2466 const __user char *uptr; /* original userland pointer */ 2467 atomic_t refcnt; 2468 struct audit_names *aname; 2469 const char iname[]; 2470 }; 2471 static_assert(offsetof(struct filename, iname) % sizeof(long) == 0); 2472 2473 static inline struct mnt_idmap *file_mnt_idmap(struct file *file) 2474 { 2475 return mnt_idmap(file->f_path.mnt); 2476 } 2477 2478 /** 2479 * is_idmapped_mnt - check whether a mount is mapped 2480 * @mnt: the mount to check 2481 * 2482 * If @mnt has an non @nop_mnt_idmap attached to it then @mnt is mapped. 2483 * 2484 * Return: true if mount is mapped, false if not. 2485 */ 2486 static inline bool is_idmapped_mnt(const struct vfsmount *mnt) 2487 { 2488 return mnt_idmap(mnt) != &nop_mnt_idmap; 2489 } 2490 2491 extern long vfs_truncate(const struct path *, loff_t); 2492 int do_truncate(struct mnt_idmap *, struct dentry *, loff_t start, 2493 unsigned int time_attrs, struct file *filp); 2494 extern int vfs_fallocate(struct file *file, int mode, loff_t offset, 2495 loff_t len); 2496 extern long do_sys_open(int dfd, const char __user *filename, int flags, 2497 umode_t mode); 2498 extern struct file *file_open_name(struct filename *, int, umode_t); 2499 extern struct file *filp_open(const char *, int, umode_t); 2500 extern struct file *file_open_root(const struct path *, 2501 const char *, int, umode_t); 2502 static inline struct file *file_open_root_mnt(struct vfsmount *mnt, 2503 const char *name, int flags, umode_t mode) 2504 { 2505 return file_open_root(&(struct path){.mnt = mnt, .dentry = mnt->mnt_root}, 2506 name, flags, mode); 2507 } 2508 struct file *dentry_open(const struct path *path, int flags, 2509 const struct cred *creds); 2510 struct file *dentry_create(const struct path *path, int flags, umode_t mode, 2511 const struct cred *cred); 2512 struct file *backing_file_open(const struct path *path, int flags, 2513 const struct path *real_path, 2514 const struct cred *cred); 2515 struct path *backing_file_real_path(struct file *f); 2516 2517 /* 2518 * file_real_path - get the path corresponding to f_inode 2519 * 2520 * When opening a backing file for a stackable filesystem (e.g., 2521 * overlayfs) f_path may be on the stackable filesystem and f_inode on 2522 * the underlying filesystem. When the path associated with f_inode is 2523 * needed, this helper should be used instead of accessing f_path 2524 * directly. 2525 */ 2526 static inline const struct path *file_real_path(struct file *f) 2527 { 2528 if (unlikely(f->f_mode & FMODE_BACKING)) 2529 return backing_file_real_path(f); 2530 return &f->f_path; 2531 } 2532 2533 static inline struct file *file_clone_open(struct file *file) 2534 { 2535 return dentry_open(&file->f_path, file->f_flags, file->f_cred); 2536 } 2537 extern int filp_close(struct file *, fl_owner_t id); 2538 2539 extern struct filename *getname_flags(const char __user *, int, int *); 2540 extern struct filename *getname_uflags(const char __user *, int); 2541 extern struct filename *getname(const char __user *); 2542 extern struct filename *getname_kernel(const char *); 2543 extern void putname(struct filename *name); 2544 2545 extern int finish_open(struct file *file, struct dentry *dentry, 2546 int (*open)(struct inode *, struct file *)); 2547 extern int finish_no_open(struct file *file, struct dentry *dentry); 2548 2549 /* Helper for the simple case when original dentry is used */ 2550 static inline int finish_open_simple(struct file *file, int error) 2551 { 2552 if (error) 2553 return error; 2554 2555 return finish_open(file, file->f_path.dentry, NULL); 2556 } 2557 2558 /* fs/dcache.c */ 2559 extern void __init vfs_caches_init_early(void); 2560 extern void __init vfs_caches_init(void); 2561 2562 extern struct kmem_cache *names_cachep; 2563 2564 #define __getname() kmem_cache_alloc(names_cachep, GFP_KERNEL) 2565 #define __putname(name) kmem_cache_free(names_cachep, (void *)(name)) 2566 2567 extern struct super_block *blockdev_superblock; 2568 static inline bool sb_is_blkdev_sb(struct super_block *sb) 2569 { 2570 return IS_ENABLED(CONFIG_BLOCK) && sb == blockdev_superblock; 2571 } 2572 2573 void emergency_thaw_all(void); 2574 extern int sync_filesystem(struct super_block *); 2575 extern const struct file_operations def_blk_fops; 2576 extern const struct file_operations def_chr_fops; 2577 2578 /* fs/char_dev.c */ 2579 #define CHRDEV_MAJOR_MAX 512 2580 /* Marks the bottom of the first segment of free char majors */ 2581 #define CHRDEV_MAJOR_DYN_END 234 2582 /* Marks the top and bottom of the second segment of free char majors */ 2583 #define CHRDEV_MAJOR_DYN_EXT_START 511 2584 #define CHRDEV_MAJOR_DYN_EXT_END 384 2585 2586 extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *); 2587 extern int register_chrdev_region(dev_t, unsigned, const char *); 2588 extern int __register_chrdev(unsigned int major, unsigned int baseminor, 2589 unsigned int count, const char *name, 2590 const struct file_operations *fops); 2591 extern void __unregister_chrdev(unsigned int major, unsigned int baseminor, 2592 unsigned int count, const char *name); 2593 extern void unregister_chrdev_region(dev_t, unsigned); 2594 extern void chrdev_show(struct seq_file *,off_t); 2595 2596 static inline int register_chrdev(unsigned int major, const char *name, 2597 const struct file_operations *fops) 2598 { 2599 return __register_chrdev(major, 0, 256, name, fops); 2600 } 2601 2602 static inline void unregister_chrdev(unsigned int major, const char *name) 2603 { 2604 __unregister_chrdev(major, 0, 256, name); 2605 } 2606 2607 extern void init_special_inode(struct inode *, umode_t, dev_t); 2608 2609 /* Invalid inode operations -- fs/bad_inode.c */ 2610 extern void make_bad_inode(struct inode *); 2611 extern bool is_bad_inode(struct inode *); 2612 2613 extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart, 2614 loff_t lend); 2615 extern int __must_check file_check_and_advance_wb_err(struct file *file); 2616 extern int __must_check file_write_and_wait_range(struct file *file, 2617 loff_t start, loff_t end); 2618 2619 static inline int file_write_and_wait(struct file *file) 2620 { 2621 return file_write_and_wait_range(file, 0, LLONG_MAX); 2622 } 2623 2624 extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end, 2625 int datasync); 2626 extern int vfs_fsync(struct file *file, int datasync); 2627 2628 extern int sync_file_range(struct file *file, loff_t offset, loff_t nbytes, 2629 unsigned int flags); 2630 2631 static inline bool iocb_is_dsync(const struct kiocb *iocb) 2632 { 2633 return (iocb->ki_flags & IOCB_DSYNC) || 2634 IS_SYNC(iocb->ki_filp->f_mapping->host); 2635 } 2636 2637 /* 2638 * Sync the bytes written if this was a synchronous write. Expect ki_pos 2639 * to already be updated for the write, and will return either the amount 2640 * of bytes passed in, or an error if syncing the file failed. 2641 */ 2642 static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count) 2643 { 2644 if (iocb_is_dsync(iocb)) { 2645 int ret = vfs_fsync_range(iocb->ki_filp, 2646 iocb->ki_pos - count, iocb->ki_pos - 1, 2647 (iocb->ki_flags & IOCB_SYNC) ? 0 : 1); 2648 if (ret) 2649 return ret; 2650 } 2651 2652 return count; 2653 } 2654 2655 extern void emergency_sync(void); 2656 extern void emergency_remount(void); 2657 2658 #ifdef CONFIG_BLOCK 2659 extern int bmap(struct inode *inode, sector_t *block); 2660 #else 2661 static inline int bmap(struct inode *inode, sector_t *block) 2662 { 2663 return -EINVAL; 2664 } 2665 #endif 2666 2667 int notify_change(struct mnt_idmap *, struct dentry *, 2668 struct iattr *, struct inode **); 2669 int inode_permission(struct mnt_idmap *, struct inode *, int); 2670 int generic_permission(struct mnt_idmap *, struct inode *, int); 2671 static inline int file_permission(struct file *file, int mask) 2672 { 2673 return inode_permission(file_mnt_idmap(file), 2674 file_inode(file), mask); 2675 } 2676 static inline int path_permission(const struct path *path, int mask) 2677 { 2678 return inode_permission(mnt_idmap(path->mnt), 2679 d_inode(path->dentry), mask); 2680 } 2681 int __check_sticky(struct mnt_idmap *idmap, struct inode *dir, 2682 struct inode *inode); 2683 2684 static inline bool execute_ok(struct inode *inode) 2685 { 2686 return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode); 2687 } 2688 2689 static inline bool inode_wrong_type(const struct inode *inode, umode_t mode) 2690 { 2691 return (inode->i_mode ^ mode) & S_IFMT; 2692 } 2693 2694 /** 2695 * file_start_write - get write access to a superblock for regular file io 2696 * @file: the file we want to write to 2697 * 2698 * This is a variant of sb_start_write() which is a noop on non-regualr file. 2699 * Should be matched with a call to file_end_write(). 2700 */ 2701 static inline void file_start_write(struct file *file) 2702 { 2703 if (!S_ISREG(file_inode(file)->i_mode)) 2704 return; 2705 sb_start_write(file_inode(file)->i_sb); 2706 } 2707 2708 static inline bool file_start_write_trylock(struct file *file) 2709 { 2710 if (!S_ISREG(file_inode(file)->i_mode)) 2711 return true; 2712 return sb_start_write_trylock(file_inode(file)->i_sb); 2713 } 2714 2715 /** 2716 * file_end_write - drop write access to a superblock of a regular file 2717 * @file: the file we wrote to 2718 * 2719 * Should be matched with a call to file_start_write(). 2720 */ 2721 static inline void file_end_write(struct file *file) 2722 { 2723 if (!S_ISREG(file_inode(file)->i_mode)) 2724 return; 2725 sb_end_write(file_inode(file)->i_sb); 2726 } 2727 2728 /** 2729 * kiocb_start_write - get write access to a superblock for async file io 2730 * @iocb: the io context we want to submit the write with 2731 * 2732 * This is a variant of sb_start_write() for async io submission. 2733 * Should be matched with a call to kiocb_end_write(). 2734 */ 2735 static inline void kiocb_start_write(struct kiocb *iocb) 2736 { 2737 struct inode *inode = file_inode(iocb->ki_filp); 2738 2739 sb_start_write(inode->i_sb); 2740 /* 2741 * Fool lockdep by telling it the lock got released so that it 2742 * doesn't complain about the held lock when we return to userspace. 2743 */ 2744 __sb_writers_release(inode->i_sb, SB_FREEZE_WRITE); 2745 } 2746 2747 /** 2748 * kiocb_end_write - drop write access to a superblock after async file io 2749 * @iocb: the io context we sumbitted the write with 2750 * 2751 * Should be matched with a call to kiocb_start_write(). 2752 */ 2753 static inline void kiocb_end_write(struct kiocb *iocb) 2754 { 2755 struct inode *inode = file_inode(iocb->ki_filp); 2756 2757 /* 2758 * Tell lockdep we inherited freeze protection from submission thread. 2759 */ 2760 __sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE); 2761 sb_end_write(inode->i_sb); 2762 } 2763 2764 /* 2765 * This is used for regular files where some users -- especially the 2766 * currently executed binary in a process, previously handled via 2767 * VM_DENYWRITE -- cannot handle concurrent write (and maybe mmap 2768 * read-write shared) accesses. 2769 * 2770 * get_write_access() gets write permission for a file. 2771 * put_write_access() releases this write permission. 2772 * deny_write_access() denies write access to a file. 2773 * allow_write_access() re-enables write access to a file. 2774 * 2775 * The i_writecount field of an inode can have the following values: 2776 * 0: no write access, no denied write access 2777 * < 0: (-i_writecount) users that denied write access to the file. 2778 * > 0: (i_writecount) users that have write access to the file. 2779 * 2780 * Normally we operate on that counter with atomic_{inc,dec} and it's safe 2781 * except for the cases where we don't hold i_writecount yet. Then we need to 2782 * use {get,deny}_write_access() - these functions check the sign and refuse 2783 * to do the change if sign is wrong. 2784 */ 2785 static inline int get_write_access(struct inode *inode) 2786 { 2787 return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY; 2788 } 2789 static inline int deny_write_access(struct file *file) 2790 { 2791 struct inode *inode = file_inode(file); 2792 return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY; 2793 } 2794 static inline void put_write_access(struct inode * inode) 2795 { 2796 atomic_dec(&inode->i_writecount); 2797 } 2798 static inline void allow_write_access(struct file *file) 2799 { 2800 if (file) 2801 atomic_inc(&file_inode(file)->i_writecount); 2802 } 2803 static inline bool inode_is_open_for_write(const struct inode *inode) 2804 { 2805 return atomic_read(&inode->i_writecount) > 0; 2806 } 2807 2808 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING) 2809 static inline void i_readcount_dec(struct inode *inode) 2810 { 2811 BUG_ON(atomic_dec_return(&inode->i_readcount) < 0); 2812 } 2813 static inline void i_readcount_inc(struct inode *inode) 2814 { 2815 atomic_inc(&inode->i_readcount); 2816 } 2817 #else 2818 static inline void i_readcount_dec(struct inode *inode) 2819 { 2820 return; 2821 } 2822 static inline void i_readcount_inc(struct inode *inode) 2823 { 2824 return; 2825 } 2826 #endif 2827 extern int do_pipe_flags(int *, int); 2828 2829 extern ssize_t kernel_read(struct file *, void *, size_t, loff_t *); 2830 ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos); 2831 extern ssize_t kernel_write(struct file *, const void *, size_t, loff_t *); 2832 extern ssize_t __kernel_write(struct file *, const void *, size_t, loff_t *); 2833 extern struct file * open_exec(const char *); 2834 2835 /* fs/dcache.c -- generic fs support functions */ 2836 extern bool is_subdir(struct dentry *, struct dentry *); 2837 extern bool path_is_under(const struct path *, const struct path *); 2838 2839 extern char *file_path(struct file *, char *, int); 2840 2841 #include <linux/err.h> 2842 2843 /* needed for stackable file system support */ 2844 extern loff_t default_llseek(struct file *file, loff_t offset, int whence); 2845 2846 extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence); 2847 2848 extern int inode_init_always(struct super_block *, struct inode *); 2849 extern void inode_init_once(struct inode *); 2850 extern void address_space_init_once(struct address_space *mapping); 2851 extern struct inode * igrab(struct inode *); 2852 extern ino_t iunique(struct super_block *, ino_t); 2853 extern int inode_needs_sync(struct inode *inode); 2854 extern int generic_delete_inode(struct inode *inode); 2855 static inline int generic_drop_inode(struct inode *inode) 2856 { 2857 return !inode->i_nlink || inode_unhashed(inode); 2858 } 2859 extern void d_mark_dontcache(struct inode *inode); 2860 2861 extern struct inode *ilookup5_nowait(struct super_block *sb, 2862 unsigned long hashval, int (*test)(struct inode *, void *), 2863 void *data); 2864 extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval, 2865 int (*test)(struct inode *, void *), void *data); 2866 extern struct inode *ilookup(struct super_block *sb, unsigned long ino); 2867 2868 extern struct inode *inode_insert5(struct inode *inode, unsigned long hashval, 2869 int (*test)(struct inode *, void *), 2870 int (*set)(struct inode *, void *), 2871 void *data); 2872 extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *); 2873 extern struct inode * iget_locked(struct super_block *, unsigned long); 2874 extern struct inode *find_inode_nowait(struct super_block *, 2875 unsigned long, 2876 int (*match)(struct inode *, 2877 unsigned long, void *), 2878 void *data); 2879 extern struct inode *find_inode_rcu(struct super_block *, unsigned long, 2880 int (*)(struct inode *, void *), void *); 2881 extern struct inode *find_inode_by_ino_rcu(struct super_block *, unsigned long); 2882 extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *); 2883 extern int insert_inode_locked(struct inode *); 2884 #ifdef CONFIG_DEBUG_LOCK_ALLOC 2885 extern void lockdep_annotate_inode_mutex_key(struct inode *inode); 2886 #else 2887 static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { }; 2888 #endif 2889 extern void unlock_new_inode(struct inode *); 2890 extern void discard_new_inode(struct inode *); 2891 extern unsigned int get_next_ino(void); 2892 extern void evict_inodes(struct super_block *sb); 2893 void dump_mapping(const struct address_space *); 2894 2895 /* 2896 * Userspace may rely on the inode number being non-zero. For example, glibc 2897 * simply ignores files with zero i_ino in unlink() and other places. 2898 * 2899 * As an additional complication, if userspace was compiled with 2900 * _FILE_OFFSET_BITS=32 on a 64-bit kernel we'll only end up reading out the 2901 * lower 32 bits, so we need to check that those aren't zero explicitly. With 2902 * _FILE_OFFSET_BITS=64, this may cause some harmless false-negatives, but 2903 * better safe than sorry. 2904 */ 2905 static inline bool is_zero_ino(ino_t ino) 2906 { 2907 return (u32)ino == 0; 2908 } 2909 2910 extern void __iget(struct inode * inode); 2911 extern void iget_failed(struct inode *); 2912 extern void clear_inode(struct inode *); 2913 extern void __destroy_inode(struct inode *); 2914 extern struct inode *new_inode_pseudo(struct super_block *sb); 2915 extern struct inode *new_inode(struct super_block *sb); 2916 extern void free_inode_nonrcu(struct inode *inode); 2917 extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *); 2918 extern int file_remove_privs(struct file *); 2919 int setattr_should_drop_sgid(struct mnt_idmap *idmap, 2920 const struct inode *inode); 2921 2922 /* 2923 * This must be used for allocating filesystems specific inodes to set 2924 * up the inode reclaim context correctly. 2925 */ 2926 static inline void * 2927 alloc_inode_sb(struct super_block *sb, struct kmem_cache *cache, gfp_t gfp) 2928 { 2929 return kmem_cache_alloc_lru(cache, &sb->s_inode_lru, gfp); 2930 } 2931 2932 extern void __insert_inode_hash(struct inode *, unsigned long hashval); 2933 static inline void insert_inode_hash(struct inode *inode) 2934 { 2935 __insert_inode_hash(inode, inode->i_ino); 2936 } 2937 2938 extern void __remove_inode_hash(struct inode *); 2939 static inline void remove_inode_hash(struct inode *inode) 2940 { 2941 if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash)) 2942 __remove_inode_hash(inode); 2943 } 2944 2945 extern void inode_sb_list_add(struct inode *inode); 2946 extern void inode_add_lru(struct inode *inode); 2947 2948 extern int sb_set_blocksize(struct super_block *, int); 2949 extern int sb_min_blocksize(struct super_block *, int); 2950 2951 extern int generic_file_mmap(struct file *, struct vm_area_struct *); 2952 extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *); 2953 extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *); 2954 int generic_write_checks_count(struct kiocb *iocb, loff_t *count); 2955 extern int generic_write_check_limits(struct file *file, loff_t pos, 2956 loff_t *count); 2957 extern int generic_file_rw_checks(struct file *file_in, struct file *file_out); 2958 ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *to, 2959 ssize_t already_read); 2960 extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *); 2961 extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *); 2962 extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *); 2963 extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *); 2964 ssize_t generic_perform_write(struct kiocb *, struct iov_iter *); 2965 ssize_t direct_write_fallback(struct kiocb *iocb, struct iov_iter *iter, 2966 ssize_t direct_written, ssize_t buffered_written); 2967 2968 ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos, 2969 rwf_t flags); 2970 ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos, 2971 rwf_t flags); 2972 ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb, 2973 struct iov_iter *iter); 2974 ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb, 2975 struct iov_iter *iter); 2976 2977 /* fs/splice.c */ 2978 ssize_t filemap_splice_read(struct file *in, loff_t *ppos, 2979 struct pipe_inode_info *pipe, 2980 size_t len, unsigned int flags); 2981 ssize_t copy_splice_read(struct file *in, loff_t *ppos, 2982 struct pipe_inode_info *pipe, 2983 size_t len, unsigned int flags); 2984 extern ssize_t iter_file_splice_write(struct pipe_inode_info *, 2985 struct file *, loff_t *, size_t, unsigned int); 2986 extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out, 2987 loff_t *opos, size_t len, unsigned int flags); 2988 2989 2990 extern void 2991 file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping); 2992 extern loff_t noop_llseek(struct file *file, loff_t offset, int whence); 2993 #define no_llseek NULL 2994 extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize); 2995 extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence); 2996 extern loff_t generic_file_llseek_size(struct file *file, loff_t offset, 2997 int whence, loff_t maxsize, loff_t eof); 2998 extern loff_t fixed_size_llseek(struct file *file, loff_t offset, 2999 int whence, loff_t size); 3000 extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t); 3001 extern loff_t no_seek_end_llseek(struct file *, loff_t, int); 3002 int rw_verify_area(int, struct file *, const loff_t *, size_t); 3003 extern int generic_file_open(struct inode * inode, struct file * filp); 3004 extern int nonseekable_open(struct inode * inode, struct file * filp); 3005 extern int stream_open(struct inode * inode, struct file * filp); 3006 3007 #ifdef CONFIG_BLOCK 3008 typedef void (dio_submit_t)(struct bio *bio, struct inode *inode, 3009 loff_t file_offset); 3010 3011 enum { 3012 /* need locking between buffered and direct access */ 3013 DIO_LOCKING = 0x01, 3014 3015 /* filesystem does not support filling holes */ 3016 DIO_SKIP_HOLES = 0x02, 3017 }; 3018 3019 ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode, 3020 struct block_device *bdev, struct iov_iter *iter, 3021 get_block_t get_block, 3022 dio_iodone_t end_io, 3023 int flags); 3024 3025 static inline ssize_t blockdev_direct_IO(struct kiocb *iocb, 3026 struct inode *inode, 3027 struct iov_iter *iter, 3028 get_block_t get_block) 3029 { 3030 return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter, 3031 get_block, NULL, DIO_LOCKING | DIO_SKIP_HOLES); 3032 } 3033 #endif 3034 3035 void inode_dio_wait(struct inode *inode); 3036 3037 /** 3038 * inode_dio_begin - signal start of a direct I/O requests 3039 * @inode: inode the direct I/O happens on 3040 * 3041 * This is called once we've finished processing a direct I/O request, 3042 * and is used to wake up callers waiting for direct I/O to be quiesced. 3043 */ 3044 static inline void inode_dio_begin(struct inode *inode) 3045 { 3046 atomic_inc(&inode->i_dio_count); 3047 } 3048 3049 /** 3050 * inode_dio_end - signal finish of a direct I/O requests 3051 * @inode: inode the direct I/O happens on 3052 * 3053 * This is called once we've finished processing a direct I/O request, 3054 * and is used to wake up callers waiting for direct I/O to be quiesced. 3055 */ 3056 static inline void inode_dio_end(struct inode *inode) 3057 { 3058 if (atomic_dec_and_test(&inode->i_dio_count)) 3059 wake_up_bit(&inode->i_state, __I_DIO_WAKEUP); 3060 } 3061 3062 extern void inode_set_flags(struct inode *inode, unsigned int flags, 3063 unsigned int mask); 3064 3065 extern const struct file_operations generic_ro_fops; 3066 3067 #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m)) 3068 3069 extern int readlink_copy(char __user *, int, const char *); 3070 extern int page_readlink(struct dentry *, char __user *, int); 3071 extern const char *page_get_link(struct dentry *, struct inode *, 3072 struct delayed_call *); 3073 extern void page_put_link(void *); 3074 extern int page_symlink(struct inode *inode, const char *symname, int len); 3075 extern const struct inode_operations page_symlink_inode_operations; 3076 extern void kfree_link(void *); 3077 void generic_fillattr(struct mnt_idmap *, u32, struct inode *, struct kstat *); 3078 void generic_fill_statx_attr(struct inode *inode, struct kstat *stat); 3079 extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int); 3080 extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int); 3081 void __inode_add_bytes(struct inode *inode, loff_t bytes); 3082 void inode_add_bytes(struct inode *inode, loff_t bytes); 3083 void __inode_sub_bytes(struct inode *inode, loff_t bytes); 3084 void inode_sub_bytes(struct inode *inode, loff_t bytes); 3085 static inline loff_t __inode_get_bytes(struct inode *inode) 3086 { 3087 return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes; 3088 } 3089 loff_t inode_get_bytes(struct inode *inode); 3090 void inode_set_bytes(struct inode *inode, loff_t bytes); 3091 const char *simple_get_link(struct dentry *, struct inode *, 3092 struct delayed_call *); 3093 extern const struct inode_operations simple_symlink_inode_operations; 3094 3095 extern int iterate_dir(struct file *, struct dir_context *); 3096 3097 int vfs_fstatat(int dfd, const char __user *filename, struct kstat *stat, 3098 int flags); 3099 int vfs_fstat(int fd, struct kstat *stat); 3100 3101 static inline int vfs_stat(const char __user *filename, struct kstat *stat) 3102 { 3103 return vfs_fstatat(AT_FDCWD, filename, stat, 0); 3104 } 3105 static inline int vfs_lstat(const char __user *name, struct kstat *stat) 3106 { 3107 return vfs_fstatat(AT_FDCWD, name, stat, AT_SYMLINK_NOFOLLOW); 3108 } 3109 3110 extern const char *vfs_get_link(struct dentry *, struct delayed_call *); 3111 extern int vfs_readlink(struct dentry *, char __user *, int); 3112 3113 extern struct file_system_type *get_filesystem(struct file_system_type *fs); 3114 extern void put_filesystem(struct file_system_type *fs); 3115 extern struct file_system_type *get_fs_type(const char *name); 3116 extern struct super_block *get_active_super(struct block_device *bdev); 3117 extern void drop_super(struct super_block *sb); 3118 extern void drop_super_exclusive(struct super_block *sb); 3119 extern void iterate_supers(void (*)(struct super_block *, void *), void *); 3120 extern void iterate_supers_type(struct file_system_type *, 3121 void (*)(struct super_block *, void *), void *); 3122 3123 extern int dcache_dir_open(struct inode *, struct file *); 3124 extern int dcache_dir_close(struct inode *, struct file *); 3125 extern loff_t dcache_dir_lseek(struct file *, loff_t, int); 3126 extern int dcache_readdir(struct file *, struct dir_context *); 3127 extern int simple_setattr(struct mnt_idmap *, struct dentry *, 3128 struct iattr *); 3129 extern int simple_getattr(struct mnt_idmap *, const struct path *, 3130 struct kstat *, u32, unsigned int); 3131 extern int simple_statfs(struct dentry *, struct kstatfs *); 3132 extern int simple_open(struct inode *inode, struct file *file); 3133 extern int simple_link(struct dentry *, struct inode *, struct dentry *); 3134 extern int simple_unlink(struct inode *, struct dentry *); 3135 extern int simple_rmdir(struct inode *, struct dentry *); 3136 void simple_rename_timestamp(struct inode *old_dir, struct dentry *old_dentry, 3137 struct inode *new_dir, struct dentry *new_dentry); 3138 extern int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry, 3139 struct inode *new_dir, struct dentry *new_dentry); 3140 extern int simple_rename(struct mnt_idmap *, struct inode *, 3141 struct dentry *, struct inode *, struct dentry *, 3142 unsigned int); 3143 extern void simple_recursive_removal(struct dentry *, 3144 void (*callback)(struct dentry *)); 3145 extern int noop_fsync(struct file *, loff_t, loff_t, int); 3146 extern ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter); 3147 extern int simple_empty(struct dentry *); 3148 extern int simple_write_begin(struct file *file, struct address_space *mapping, 3149 loff_t pos, unsigned len, 3150 struct page **pagep, void **fsdata); 3151 extern const struct address_space_operations ram_aops; 3152 extern int always_delete_dentry(const struct dentry *); 3153 extern struct inode *alloc_anon_inode(struct super_block *); 3154 extern int simple_nosetlease(struct file *, int, struct file_lock **, void **); 3155 extern const struct dentry_operations simple_dentry_operations; 3156 3157 extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags); 3158 extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *); 3159 extern const struct file_operations simple_dir_operations; 3160 extern const struct inode_operations simple_dir_inode_operations; 3161 extern void make_empty_dir_inode(struct inode *inode); 3162 extern bool is_empty_dir_inode(struct inode *inode); 3163 struct tree_descr { const char *name; const struct file_operations *ops; int mode; }; 3164 struct dentry *d_alloc_name(struct dentry *, const char *); 3165 extern int simple_fill_super(struct super_block *, unsigned long, 3166 const struct tree_descr *); 3167 extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count); 3168 extern void simple_release_fs(struct vfsmount **mount, int *count); 3169 3170 extern ssize_t simple_read_from_buffer(void __user *to, size_t count, 3171 loff_t *ppos, const void *from, size_t available); 3172 extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos, 3173 const void __user *from, size_t count); 3174 3175 struct offset_ctx { 3176 struct xarray xa; 3177 u32 next_offset; 3178 }; 3179 3180 void simple_offset_init(struct offset_ctx *octx); 3181 int simple_offset_add(struct offset_ctx *octx, struct dentry *dentry); 3182 void simple_offset_remove(struct offset_ctx *octx, struct dentry *dentry); 3183 int simple_offset_rename_exchange(struct inode *old_dir, 3184 struct dentry *old_dentry, 3185 struct inode *new_dir, 3186 struct dentry *new_dentry); 3187 void simple_offset_destroy(struct offset_ctx *octx); 3188 3189 extern const struct file_operations simple_offset_dir_operations; 3190 3191 extern int __generic_file_fsync(struct file *, loff_t, loff_t, int); 3192 extern int generic_file_fsync(struct file *, loff_t, loff_t, int); 3193 3194 extern int generic_check_addressable(unsigned, u64); 3195 3196 extern void generic_set_encrypted_ci_d_ops(struct dentry *dentry); 3197 3198 static inline bool sb_has_encoding(const struct super_block *sb) 3199 { 3200 #if IS_ENABLED(CONFIG_UNICODE) 3201 return !!sb->s_encoding; 3202 #else 3203 return false; 3204 #endif 3205 } 3206 3207 int may_setattr(struct mnt_idmap *idmap, struct inode *inode, 3208 unsigned int ia_valid); 3209 int setattr_prepare(struct mnt_idmap *, struct dentry *, struct iattr *); 3210 extern int inode_newsize_ok(const struct inode *, loff_t offset); 3211 void setattr_copy(struct mnt_idmap *, struct inode *inode, 3212 const struct iattr *attr); 3213 3214 extern int file_update_time(struct file *file); 3215 3216 static inline bool vma_is_dax(const struct vm_area_struct *vma) 3217 { 3218 return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host); 3219 } 3220 3221 static inline bool vma_is_fsdax(struct vm_area_struct *vma) 3222 { 3223 struct inode *inode; 3224 3225 if (!IS_ENABLED(CONFIG_FS_DAX) || !vma->vm_file) 3226 return false; 3227 if (!vma_is_dax(vma)) 3228 return false; 3229 inode = file_inode(vma->vm_file); 3230 if (S_ISCHR(inode->i_mode)) 3231 return false; /* device-dax */ 3232 return true; 3233 } 3234 3235 static inline int iocb_flags(struct file *file) 3236 { 3237 int res = 0; 3238 if (file->f_flags & O_APPEND) 3239 res |= IOCB_APPEND; 3240 if (file->f_flags & O_DIRECT) 3241 res |= IOCB_DIRECT; 3242 if (file->f_flags & O_DSYNC) 3243 res |= IOCB_DSYNC; 3244 if (file->f_flags & __O_SYNC) 3245 res |= IOCB_SYNC; 3246 return res; 3247 } 3248 3249 static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags) 3250 { 3251 int kiocb_flags = 0; 3252 3253 /* make sure there's no overlap between RWF and private IOCB flags */ 3254 BUILD_BUG_ON((__force int) RWF_SUPPORTED & IOCB_EVENTFD); 3255 3256 if (!flags) 3257 return 0; 3258 if (unlikely(flags & ~RWF_SUPPORTED)) 3259 return -EOPNOTSUPP; 3260 3261 if (flags & RWF_NOWAIT) { 3262 if (!(ki->ki_filp->f_mode & FMODE_NOWAIT)) 3263 return -EOPNOTSUPP; 3264 kiocb_flags |= IOCB_NOIO; 3265 } 3266 kiocb_flags |= (__force int) (flags & RWF_SUPPORTED); 3267 if (flags & RWF_SYNC) 3268 kiocb_flags |= IOCB_DSYNC; 3269 3270 ki->ki_flags |= kiocb_flags; 3271 return 0; 3272 } 3273 3274 static inline ino_t parent_ino(struct dentry *dentry) 3275 { 3276 ino_t res; 3277 3278 /* 3279 * Don't strictly need d_lock here? If the parent ino could change 3280 * then surely we'd have a deeper race in the caller? 3281 */ 3282 spin_lock(&dentry->d_lock); 3283 res = dentry->d_parent->d_inode->i_ino; 3284 spin_unlock(&dentry->d_lock); 3285 return res; 3286 } 3287 3288 /* Transaction based IO helpers */ 3289 3290 /* 3291 * An argresp is stored in an allocated page and holds the 3292 * size of the argument or response, along with its content 3293 */ 3294 struct simple_transaction_argresp { 3295 ssize_t size; 3296 char data[]; 3297 }; 3298 3299 #define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp)) 3300 3301 char *simple_transaction_get(struct file *file, const char __user *buf, 3302 size_t size); 3303 ssize_t simple_transaction_read(struct file *file, char __user *buf, 3304 size_t size, loff_t *pos); 3305 int simple_transaction_release(struct inode *inode, struct file *file); 3306 3307 void simple_transaction_set(struct file *file, size_t n); 3308 3309 /* 3310 * simple attribute files 3311 * 3312 * These attributes behave similar to those in sysfs: 3313 * 3314 * Writing to an attribute immediately sets a value, an open file can be 3315 * written to multiple times. 3316 * 3317 * Reading from an attribute creates a buffer from the value that might get 3318 * read with multiple read calls. When the attribute has been read 3319 * completely, no further read calls are possible until the file is opened 3320 * again. 3321 * 3322 * All attributes contain a text representation of a numeric value 3323 * that are accessed with the get() and set() functions. 3324 */ 3325 #define DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, __is_signed) \ 3326 static int __fops ## _open(struct inode *inode, struct file *file) \ 3327 { \ 3328 __simple_attr_check_format(__fmt, 0ull); \ 3329 return simple_attr_open(inode, file, __get, __set, __fmt); \ 3330 } \ 3331 static const struct file_operations __fops = { \ 3332 .owner = THIS_MODULE, \ 3333 .open = __fops ## _open, \ 3334 .release = simple_attr_release, \ 3335 .read = simple_attr_read, \ 3336 .write = (__is_signed) ? simple_attr_write_signed : simple_attr_write, \ 3337 .llseek = generic_file_llseek, \ 3338 } 3339 3340 #define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt) \ 3341 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, false) 3342 3343 #define DEFINE_SIMPLE_ATTRIBUTE_SIGNED(__fops, __get, __set, __fmt) \ 3344 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, true) 3345 3346 static inline __printf(1, 2) 3347 void __simple_attr_check_format(const char *fmt, ...) 3348 { 3349 /* don't do anything, just let the compiler check the arguments; */ 3350 } 3351 3352 int simple_attr_open(struct inode *inode, struct file *file, 3353 int (*get)(void *, u64 *), int (*set)(void *, u64), 3354 const char *fmt); 3355 int simple_attr_release(struct inode *inode, struct file *file); 3356 ssize_t simple_attr_read(struct file *file, char __user *buf, 3357 size_t len, loff_t *ppos); 3358 ssize_t simple_attr_write(struct file *file, const char __user *buf, 3359 size_t len, loff_t *ppos); 3360 ssize_t simple_attr_write_signed(struct file *file, const char __user *buf, 3361 size_t len, loff_t *ppos); 3362 3363 struct ctl_table; 3364 int __init list_bdev_fs_names(char *buf, size_t size); 3365 3366 #define __FMODE_EXEC ((__force int) FMODE_EXEC) 3367 #define __FMODE_NONOTIFY ((__force int) FMODE_NONOTIFY) 3368 3369 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE]) 3370 #define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \ 3371 (flag & __FMODE_NONOTIFY))) 3372 3373 static inline bool is_sxid(umode_t mode) 3374 { 3375 return mode & (S_ISUID | S_ISGID); 3376 } 3377 3378 static inline int check_sticky(struct mnt_idmap *idmap, 3379 struct inode *dir, struct inode *inode) 3380 { 3381 if (!(dir->i_mode & S_ISVTX)) 3382 return 0; 3383 3384 return __check_sticky(idmap, dir, inode); 3385 } 3386 3387 static inline void inode_has_no_xattr(struct inode *inode) 3388 { 3389 if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC)) 3390 inode->i_flags |= S_NOSEC; 3391 } 3392 3393 static inline bool is_root_inode(struct inode *inode) 3394 { 3395 return inode == inode->i_sb->s_root->d_inode; 3396 } 3397 3398 static inline bool dir_emit(struct dir_context *ctx, 3399 const char *name, int namelen, 3400 u64 ino, unsigned type) 3401 { 3402 return ctx->actor(ctx, name, namelen, ctx->pos, ino, type); 3403 } 3404 static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx) 3405 { 3406 return ctx->actor(ctx, ".", 1, ctx->pos, 3407 file->f_path.dentry->d_inode->i_ino, DT_DIR); 3408 } 3409 static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx) 3410 { 3411 return ctx->actor(ctx, "..", 2, ctx->pos, 3412 parent_ino(file->f_path.dentry), DT_DIR); 3413 } 3414 static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx) 3415 { 3416 if (ctx->pos == 0) { 3417 if (!dir_emit_dot(file, ctx)) 3418 return false; 3419 ctx->pos = 1; 3420 } 3421 if (ctx->pos == 1) { 3422 if (!dir_emit_dotdot(file, ctx)) 3423 return false; 3424 ctx->pos = 2; 3425 } 3426 return true; 3427 } 3428 static inline bool dir_relax(struct inode *inode) 3429 { 3430 inode_unlock(inode); 3431 inode_lock(inode); 3432 return !IS_DEADDIR(inode); 3433 } 3434 3435 static inline bool dir_relax_shared(struct inode *inode) 3436 { 3437 inode_unlock_shared(inode); 3438 inode_lock_shared(inode); 3439 return !IS_DEADDIR(inode); 3440 } 3441 3442 extern bool path_noexec(const struct path *path); 3443 extern void inode_nohighmem(struct inode *inode); 3444 3445 /* mm/fadvise.c */ 3446 extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len, 3447 int advice); 3448 extern int generic_fadvise(struct file *file, loff_t offset, loff_t len, 3449 int advice); 3450 3451 #endif /* _LINUX_FS_H */ 3452