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