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 #include <linux/maple_tree.h> 47 #include <linux/rw_hint.h> 48 49 #include <asm/byteorder.h> 50 #include <uapi/linux/fs.h> 51 52 struct backing_dev_info; 53 struct bdi_writeback; 54 struct bio; 55 struct io_comp_batch; 56 struct export_operations; 57 struct fiemap_extent_info; 58 struct hd_geometry; 59 struct iovec; 60 struct kiocb; 61 struct kobject; 62 struct pipe_inode_info; 63 struct poll_table_struct; 64 struct kstatfs; 65 struct vm_area_struct; 66 struct vfsmount; 67 struct cred; 68 struct swap_info_struct; 69 struct seq_file; 70 struct workqueue_struct; 71 struct iov_iter; 72 struct fscrypt_inode_info; 73 struct fscrypt_operations; 74 struct fsverity_info; 75 struct fsverity_operations; 76 struct fs_context; 77 struct fs_parameter_spec; 78 struct fileattr; 79 struct iomap_ops; 80 81 extern void __init inode_init(void); 82 extern void __init inode_init_early(void); 83 extern void __init files_init(void); 84 extern void __init files_maxfiles_init(void); 85 86 extern unsigned long get_max_files(void); 87 extern unsigned int sysctl_nr_open; 88 89 typedef __kernel_rwf_t rwf_t; 90 91 struct buffer_head; 92 typedef int (get_block_t)(struct inode *inode, sector_t iblock, 93 struct buffer_head *bh_result, int create); 94 typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset, 95 ssize_t bytes, void *private); 96 97 #define MAY_EXEC 0x00000001 98 #define MAY_WRITE 0x00000002 99 #define MAY_READ 0x00000004 100 #define MAY_APPEND 0x00000008 101 #define MAY_ACCESS 0x00000010 102 #define MAY_OPEN 0x00000020 103 #define MAY_CHDIR 0x00000040 104 /* called from RCU mode, don't block */ 105 #define MAY_NOT_BLOCK 0x00000080 106 107 /* 108 * flags in file.f_mode. Note that FMODE_READ and FMODE_WRITE must correspond 109 * to O_WRONLY and O_RDWR via the strange trick in do_dentry_open() 110 */ 111 112 /* file is open for reading */ 113 #define FMODE_READ ((__force fmode_t)0x1) 114 /* file is open for writing */ 115 #define FMODE_WRITE ((__force fmode_t)0x2) 116 /* file is seekable */ 117 #define FMODE_LSEEK ((__force fmode_t)0x4) 118 /* file can be accessed using pread */ 119 #define FMODE_PREAD ((__force fmode_t)0x8) 120 /* file can be accessed using pwrite */ 121 #define FMODE_PWRITE ((__force fmode_t)0x10) 122 /* File is opened for execution with sys_execve / sys_uselib */ 123 #define FMODE_EXEC ((__force fmode_t)0x20) 124 /* File writes are restricted (block device specific) */ 125 #define FMODE_WRITE_RESTRICTED ((__force fmode_t)0x40) 126 /* 32bit hashes as llseek() offset (for directories) */ 127 #define FMODE_32BITHASH ((__force fmode_t)0x200) 128 /* 64bit hashes as llseek() offset (for directories) */ 129 #define FMODE_64BITHASH ((__force fmode_t)0x400) 130 131 /* 132 * Don't update ctime and mtime. 133 * 134 * Currently a special hack for the XFS open_by_handle ioctl, but we'll 135 * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon. 136 */ 137 #define FMODE_NOCMTIME ((__force fmode_t)0x800) 138 139 /* Expect random access pattern */ 140 #define FMODE_RANDOM ((__force fmode_t)0x1000) 141 142 /* File is huge (eg. /dev/mem): treat loff_t as unsigned */ 143 #define FMODE_UNSIGNED_OFFSET ((__force fmode_t)0x2000) 144 145 /* File is opened with O_PATH; almost nothing can be done with it */ 146 #define FMODE_PATH ((__force fmode_t)0x4000) 147 148 /* File needs atomic accesses to f_pos */ 149 #define FMODE_ATOMIC_POS ((__force fmode_t)0x8000) 150 /* Write access to underlying fs */ 151 #define FMODE_WRITER ((__force fmode_t)0x10000) 152 /* Has read method(s) */ 153 #define FMODE_CAN_READ ((__force fmode_t)0x20000) 154 /* Has write method(s) */ 155 #define FMODE_CAN_WRITE ((__force fmode_t)0x40000) 156 157 #define FMODE_OPENED ((__force fmode_t)0x80000) 158 #define FMODE_CREATED ((__force fmode_t)0x100000) 159 160 /* File is stream-like */ 161 #define FMODE_STREAM ((__force fmode_t)0x200000) 162 163 /* File supports DIRECT IO */ 164 #define FMODE_CAN_ODIRECT ((__force fmode_t)0x400000) 165 166 #define FMODE_NOREUSE ((__force fmode_t)0x800000) 167 168 /* File supports non-exclusive O_DIRECT writes from multiple threads */ 169 #define FMODE_DIO_PARALLEL_WRITE ((__force fmode_t)0x1000000) 170 171 /* File is embedded in backing_file object */ 172 #define FMODE_BACKING ((__force fmode_t)0x2000000) 173 174 /* File was opened by fanotify and shouldn't generate fanotify events */ 175 #define FMODE_NONOTIFY ((__force fmode_t)0x4000000) 176 177 /* File is capable of returning -EAGAIN if I/O will block */ 178 #define FMODE_NOWAIT ((__force fmode_t)0x8000000) 179 180 /* File represents mount that needs unmounting */ 181 #define FMODE_NEED_UNMOUNT ((__force fmode_t)0x10000000) 182 183 /* File does not contribute to nr_files count */ 184 #define FMODE_NOACCOUNT ((__force fmode_t)0x20000000) 185 186 /* File supports async buffered reads */ 187 #define FMODE_BUF_RASYNC ((__force fmode_t)0x40000000) 188 189 /* File supports async nowait buffered writes */ 190 #define FMODE_BUF_WASYNC ((__force fmode_t)0x80000000) 191 192 /* 193 * Attribute flags. These should be or-ed together to figure out what 194 * has been changed! 195 */ 196 #define ATTR_MODE (1 << 0) 197 #define ATTR_UID (1 << 1) 198 #define ATTR_GID (1 << 2) 199 #define ATTR_SIZE (1 << 3) 200 #define ATTR_ATIME (1 << 4) 201 #define ATTR_MTIME (1 << 5) 202 #define ATTR_CTIME (1 << 6) 203 #define ATTR_ATIME_SET (1 << 7) 204 #define ATTR_MTIME_SET (1 << 8) 205 #define ATTR_FORCE (1 << 9) /* Not a change, but a change it */ 206 #define ATTR_KILL_SUID (1 << 11) 207 #define ATTR_KILL_SGID (1 << 12) 208 #define ATTR_FILE (1 << 13) 209 #define ATTR_KILL_PRIV (1 << 14) 210 #define ATTR_OPEN (1 << 15) /* Truncating from open(O_TRUNC) */ 211 #define ATTR_TIMES_SET (1 << 16) 212 #define ATTR_TOUCH (1 << 17) 213 214 /* 215 * Whiteout is represented by a char device. The following constants define the 216 * mode and device number to use. 217 */ 218 #define WHITEOUT_MODE 0 219 #define WHITEOUT_DEV 0 220 221 /* 222 * This is the Inode Attributes structure, used for notify_change(). It 223 * uses the above definitions as flags, to know which values have changed. 224 * Also, in this manner, a Filesystem can look at only the values it cares 225 * about. Basically, these are the attributes that the VFS layer can 226 * request to change from the FS layer. 227 * 228 * Derek Atkins <warlord@MIT.EDU> 94-10-20 229 */ 230 struct iattr { 231 unsigned int ia_valid; 232 umode_t ia_mode; 233 /* 234 * The two anonymous unions wrap structures with the same member. 235 * 236 * Filesystems raising FS_ALLOW_IDMAP need to use ia_vfs{g,u}id which 237 * are a dedicated type requiring the filesystem to use the dedicated 238 * helpers. Other filesystem can continue to use ia_{g,u}id until they 239 * have been ported. 240 * 241 * They always contain the same value. In other words FS_ALLOW_IDMAP 242 * pass down the same value on idmapped mounts as they would on regular 243 * mounts. 244 */ 245 union { 246 kuid_t ia_uid; 247 vfsuid_t ia_vfsuid; 248 }; 249 union { 250 kgid_t ia_gid; 251 vfsgid_t ia_vfsgid; 252 }; 253 loff_t ia_size; 254 struct timespec64 ia_atime; 255 struct timespec64 ia_mtime; 256 struct timespec64 ia_ctime; 257 258 /* 259 * Not an attribute, but an auxiliary info for filesystems wanting to 260 * implement an ftruncate() like method. NOTE: filesystem should 261 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL). 262 */ 263 struct file *ia_file; 264 }; 265 266 /* 267 * Includes for diskquotas. 268 */ 269 #include <linux/quota.h> 270 271 /* 272 * Maximum number of layers of fs stack. Needs to be limited to 273 * prevent kernel stack overflow 274 */ 275 #define FILESYSTEM_MAX_STACK_DEPTH 2 276 277 /** 278 * enum positive_aop_returns - aop return codes with specific semantics 279 * 280 * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has 281 * completed, that the page is still locked, and 282 * should be considered active. The VM uses this hint 283 * to return the page to the active list -- it won't 284 * be a candidate for writeback again in the near 285 * future. Other callers must be careful to unlock 286 * the page if they get this return. Returned by 287 * writepage(); 288 * 289 * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has 290 * unlocked it and the page might have been truncated. 291 * The caller should back up to acquiring a new page and 292 * trying again. The aop will be taking reasonable 293 * precautions not to livelock. If the caller held a page 294 * reference, it should drop it before retrying. Returned 295 * by read_folio(). 296 * 297 * address_space_operation functions return these large constants to indicate 298 * special semantics to the caller. These are much larger than the bytes in a 299 * page to allow for functions that return the number of bytes operated on in a 300 * given page. 301 */ 302 303 enum positive_aop_returns { 304 AOP_WRITEPAGE_ACTIVATE = 0x80000, 305 AOP_TRUNCATED_PAGE = 0x80001, 306 }; 307 308 /* 309 * oh the beauties of C type declarations. 310 */ 311 struct page; 312 struct address_space; 313 struct writeback_control; 314 struct readahead_control; 315 316 /* Match RWF_* bits to IOCB bits */ 317 #define IOCB_HIPRI (__force int) RWF_HIPRI 318 #define IOCB_DSYNC (__force int) RWF_DSYNC 319 #define IOCB_SYNC (__force int) RWF_SYNC 320 #define IOCB_NOWAIT (__force int) RWF_NOWAIT 321 #define IOCB_APPEND (__force int) RWF_APPEND 322 323 /* non-RWF related bits - start at 16 */ 324 #define IOCB_EVENTFD (1 << 16) 325 #define IOCB_DIRECT (1 << 17) 326 #define IOCB_WRITE (1 << 18) 327 /* iocb->ki_waitq is valid */ 328 #define IOCB_WAITQ (1 << 19) 329 #define IOCB_NOIO (1 << 20) 330 /* can use bio alloc cache */ 331 #define IOCB_ALLOC_CACHE (1 << 21) 332 /* 333 * IOCB_DIO_CALLER_COMP can be set by the iocb owner, to indicate that the 334 * iocb completion can be passed back to the owner for execution from a safe 335 * context rather than needing to be punted through a workqueue. If this 336 * flag is set, the bio completion handling may set iocb->dio_complete to a 337 * handler function and iocb->private to context information for that handler. 338 * The issuer should call the handler with that context information from task 339 * context to complete the processing of the iocb. Note that while this 340 * provides a task context for the dio_complete() callback, it should only be 341 * used on the completion side for non-IO generating completions. It's fine to 342 * call blocking functions from this callback, but they should not wait for 343 * unrelated IO (like cache flushing, new IO generation, etc). 344 */ 345 #define IOCB_DIO_CALLER_COMP (1 << 22) 346 /* kiocb is a read or write operation submitted by fs/aio.c. */ 347 #define IOCB_AIO_RW (1 << 23) 348 349 /* for use in trace events */ 350 #define TRACE_IOCB_STRINGS \ 351 { IOCB_HIPRI, "HIPRI" }, \ 352 { IOCB_DSYNC, "DSYNC" }, \ 353 { IOCB_SYNC, "SYNC" }, \ 354 { IOCB_NOWAIT, "NOWAIT" }, \ 355 { IOCB_APPEND, "APPEND" }, \ 356 { IOCB_EVENTFD, "EVENTFD"}, \ 357 { IOCB_DIRECT, "DIRECT" }, \ 358 { IOCB_WRITE, "WRITE" }, \ 359 { IOCB_WAITQ, "WAITQ" }, \ 360 { IOCB_NOIO, "NOIO" }, \ 361 { IOCB_ALLOC_CACHE, "ALLOC_CACHE" }, \ 362 { IOCB_DIO_CALLER_COMP, "CALLER_COMP" } 363 364 struct kiocb { 365 struct file *ki_filp; 366 loff_t ki_pos; 367 void (*ki_complete)(struct kiocb *iocb, long ret); 368 void *private; 369 int ki_flags; 370 u16 ki_ioprio; /* See linux/ioprio.h */ 371 union { 372 /* 373 * Only used for async buffered reads, where it denotes the 374 * page waitqueue associated with completing the read. Valid 375 * IFF IOCB_WAITQ is set. 376 */ 377 struct wait_page_queue *ki_waitq; 378 /* 379 * Can be used for O_DIRECT IO, where the completion handling 380 * is punted back to the issuer of the IO. May only be set 381 * if IOCB_DIO_CALLER_COMP is set by the issuer, and the issuer 382 * must then check for presence of this handler when ki_complete 383 * is invoked. The data passed in to this handler must be 384 * assigned to ->private when dio_complete is assigned. 385 */ 386 ssize_t (*dio_complete)(void *data); 387 }; 388 }; 389 390 static inline bool is_sync_kiocb(struct kiocb *kiocb) 391 { 392 return kiocb->ki_complete == NULL; 393 } 394 395 struct address_space_operations { 396 int (*writepage)(struct page *page, struct writeback_control *wbc); 397 int (*read_folio)(struct file *, struct folio *); 398 399 /* Write back some dirty pages from this mapping. */ 400 int (*writepages)(struct address_space *, struct writeback_control *); 401 402 /* Mark a folio dirty. Return true if this dirtied it */ 403 bool (*dirty_folio)(struct address_space *, struct folio *); 404 405 void (*readahead)(struct readahead_control *); 406 407 int (*write_begin)(struct file *, struct address_space *mapping, 408 loff_t pos, unsigned len, 409 struct page **pagep, void **fsdata); 410 int (*write_end)(struct file *, struct address_space *mapping, 411 loff_t pos, unsigned len, unsigned copied, 412 struct page *page, void *fsdata); 413 414 /* Unfortunately this kludge is needed for FIBMAP. Don't use it */ 415 sector_t (*bmap)(struct address_space *, sector_t); 416 void (*invalidate_folio) (struct folio *, size_t offset, size_t len); 417 bool (*release_folio)(struct folio *, gfp_t); 418 void (*free_folio)(struct folio *folio); 419 ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter); 420 /* 421 * migrate the contents of a folio to the specified target. If 422 * migrate_mode is MIGRATE_ASYNC, it must not block. 423 */ 424 int (*migrate_folio)(struct address_space *, struct folio *dst, 425 struct folio *src, enum migrate_mode); 426 int (*launder_folio)(struct folio *); 427 bool (*is_partially_uptodate) (struct folio *, size_t from, 428 size_t count); 429 void (*is_dirty_writeback) (struct folio *, bool *dirty, bool *wb); 430 int (*error_remove_folio)(struct address_space *, struct folio *); 431 432 /* swapfile support */ 433 int (*swap_activate)(struct swap_info_struct *sis, struct file *file, 434 sector_t *span); 435 void (*swap_deactivate)(struct file *file); 436 int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter); 437 }; 438 439 extern const struct address_space_operations empty_aops; 440 441 /** 442 * struct address_space - Contents of a cacheable, mappable object. 443 * @host: Owner, either the inode or the block_device. 444 * @i_pages: Cached pages. 445 * @invalidate_lock: Guards coherency between page cache contents and 446 * file offset->disk block mappings in the filesystem during invalidates. 447 * It is also used to block modification of page cache contents through 448 * memory mappings. 449 * @gfp_mask: Memory allocation flags to use for allocating pages. 450 * @i_mmap_writable: Number of VM_SHARED, VM_MAYWRITE mappings. 451 * @nr_thps: Number of THPs in the pagecache (non-shmem only). 452 * @i_mmap: Tree of private and shared mappings. 453 * @i_mmap_rwsem: Protects @i_mmap and @i_mmap_writable. 454 * @nrpages: Number of page entries, protected by the i_pages lock. 455 * @writeback_index: Writeback starts here. 456 * @a_ops: Methods. 457 * @flags: Error bits and flags (AS_*). 458 * @wb_err: The most recent error which has occurred. 459 * @i_private_lock: For use by the owner of the address_space. 460 * @i_private_list: For use by the owner of the address_space. 461 * @i_private_data: For use by the owner of the address_space. 462 */ 463 struct address_space { 464 struct inode *host; 465 struct xarray i_pages; 466 struct rw_semaphore invalidate_lock; 467 gfp_t gfp_mask; 468 atomic_t i_mmap_writable; 469 #ifdef CONFIG_READ_ONLY_THP_FOR_FS 470 /* number of thp, only for non-shmem files */ 471 atomic_t nr_thps; 472 #endif 473 struct rb_root_cached i_mmap; 474 unsigned long nrpages; 475 pgoff_t writeback_index; 476 const struct address_space_operations *a_ops; 477 unsigned long flags; 478 errseq_t wb_err; 479 spinlock_t i_private_lock; 480 struct list_head i_private_list; 481 struct rw_semaphore i_mmap_rwsem; 482 void * i_private_data; 483 } __attribute__((aligned(sizeof(long)))) __randomize_layout; 484 /* 485 * On most architectures that alignment is already the case; but 486 * must be enforced here for CRIS, to let the least significant bit 487 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON. 488 */ 489 490 /* XArray tags, for tagging dirty and writeback pages in the pagecache. */ 491 #define PAGECACHE_TAG_DIRTY XA_MARK_0 492 #define PAGECACHE_TAG_WRITEBACK XA_MARK_1 493 #define PAGECACHE_TAG_TOWRITE XA_MARK_2 494 495 /* 496 * Returns true if any of the pages in the mapping are marked with the tag. 497 */ 498 static inline bool mapping_tagged(struct address_space *mapping, xa_mark_t tag) 499 { 500 return xa_marked(&mapping->i_pages, tag); 501 } 502 503 static inline void i_mmap_lock_write(struct address_space *mapping) 504 { 505 down_write(&mapping->i_mmap_rwsem); 506 } 507 508 static inline int i_mmap_trylock_write(struct address_space *mapping) 509 { 510 return down_write_trylock(&mapping->i_mmap_rwsem); 511 } 512 513 static inline void i_mmap_unlock_write(struct address_space *mapping) 514 { 515 up_write(&mapping->i_mmap_rwsem); 516 } 517 518 static inline int i_mmap_trylock_read(struct address_space *mapping) 519 { 520 return down_read_trylock(&mapping->i_mmap_rwsem); 521 } 522 523 static inline void i_mmap_lock_read(struct address_space *mapping) 524 { 525 down_read(&mapping->i_mmap_rwsem); 526 } 527 528 static inline void i_mmap_unlock_read(struct address_space *mapping) 529 { 530 up_read(&mapping->i_mmap_rwsem); 531 } 532 533 static inline void i_mmap_assert_locked(struct address_space *mapping) 534 { 535 lockdep_assert_held(&mapping->i_mmap_rwsem); 536 } 537 538 static inline void i_mmap_assert_write_locked(struct address_space *mapping) 539 { 540 lockdep_assert_held_write(&mapping->i_mmap_rwsem); 541 } 542 543 /* 544 * Might pages of this file be mapped into userspace? 545 */ 546 static inline int mapping_mapped(struct address_space *mapping) 547 { 548 return !RB_EMPTY_ROOT(&mapping->i_mmap.rb_root); 549 } 550 551 /* 552 * Might pages of this file have been modified in userspace? 553 * Note that i_mmap_writable counts all VM_SHARED, VM_MAYWRITE vmas: do_mmap 554 * marks vma as VM_SHARED if it is shared, and the file was opened for 555 * writing i.e. vma may be mprotected writable even if now readonly. 556 * 557 * If i_mmap_writable is negative, no new writable mappings are allowed. You 558 * can only deny writable mappings, if none exists right now. 559 */ 560 static inline int mapping_writably_mapped(struct address_space *mapping) 561 { 562 return atomic_read(&mapping->i_mmap_writable) > 0; 563 } 564 565 static inline int mapping_map_writable(struct address_space *mapping) 566 { 567 return atomic_inc_unless_negative(&mapping->i_mmap_writable) ? 568 0 : -EPERM; 569 } 570 571 static inline void mapping_unmap_writable(struct address_space *mapping) 572 { 573 atomic_dec(&mapping->i_mmap_writable); 574 } 575 576 static inline int mapping_deny_writable(struct address_space *mapping) 577 { 578 return atomic_dec_unless_positive(&mapping->i_mmap_writable) ? 579 0 : -EBUSY; 580 } 581 582 static inline void mapping_allow_writable(struct address_space *mapping) 583 { 584 atomic_inc(&mapping->i_mmap_writable); 585 } 586 587 /* 588 * Use sequence counter to get consistent i_size on 32-bit processors. 589 */ 590 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 591 #include <linux/seqlock.h> 592 #define __NEED_I_SIZE_ORDERED 593 #define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount) 594 #else 595 #define i_size_ordered_init(inode) do { } while (0) 596 #endif 597 598 struct posix_acl; 599 #define ACL_NOT_CACHED ((void *)(-1)) 600 /* 601 * ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to 602 * cache the ACL. This also means that ->get_inode_acl() can be called in RCU 603 * mode with the LOOKUP_RCU flag. 604 */ 605 #define ACL_DONT_CACHE ((void *)(-3)) 606 607 static inline struct posix_acl * 608 uncached_acl_sentinel(struct task_struct *task) 609 { 610 return (void *)task + 1; 611 } 612 613 static inline bool 614 is_uncached_acl(struct posix_acl *acl) 615 { 616 return (long)acl & 1; 617 } 618 619 #define IOP_FASTPERM 0x0001 620 #define IOP_LOOKUP 0x0002 621 #define IOP_NOFOLLOW 0x0004 622 #define IOP_XATTR 0x0008 623 #define IOP_DEFAULT_READLINK 0x0010 624 625 struct fsnotify_mark_connector; 626 627 /* 628 * Keep mostly read-only and often accessed (especially for 629 * the RCU path lookup and 'stat' data) fields at the beginning 630 * of the 'struct inode' 631 */ 632 struct inode { 633 umode_t i_mode; 634 unsigned short i_opflags; 635 kuid_t i_uid; 636 kgid_t i_gid; 637 unsigned int i_flags; 638 639 #ifdef CONFIG_FS_POSIX_ACL 640 struct posix_acl *i_acl; 641 struct posix_acl *i_default_acl; 642 #endif 643 644 const struct inode_operations *i_op; 645 struct super_block *i_sb; 646 struct address_space *i_mapping; 647 648 #ifdef CONFIG_SECURITY 649 void *i_security; 650 #endif 651 652 /* Stat data, not accessed from path walking */ 653 unsigned long i_ino; 654 /* 655 * Filesystems may only read i_nlink directly. They shall use the 656 * following functions for modification: 657 * 658 * (set|clear|inc|drop)_nlink 659 * inode_(inc|dec)_link_count 660 */ 661 union { 662 const unsigned int i_nlink; 663 unsigned int __i_nlink; 664 }; 665 dev_t i_rdev; 666 loff_t i_size; 667 struct timespec64 __i_atime; 668 struct timespec64 __i_mtime; 669 struct timespec64 __i_ctime; /* use inode_*_ctime accessors! */ 670 spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */ 671 unsigned short i_bytes; 672 u8 i_blkbits; 673 enum rw_hint i_write_hint; 674 blkcnt_t i_blocks; 675 676 #ifdef __NEED_I_SIZE_ORDERED 677 seqcount_t i_size_seqcount; 678 #endif 679 680 /* Misc */ 681 unsigned long i_state; 682 struct rw_semaphore i_rwsem; 683 684 unsigned long dirtied_when; /* jiffies of first dirtying */ 685 unsigned long dirtied_time_when; 686 687 struct hlist_node i_hash; 688 struct list_head i_io_list; /* backing dev IO list */ 689 #ifdef CONFIG_CGROUP_WRITEBACK 690 struct bdi_writeback *i_wb; /* the associated cgroup wb */ 691 692 /* foreign inode detection, see wbc_detach_inode() */ 693 int i_wb_frn_winner; 694 u16 i_wb_frn_avg_time; 695 u16 i_wb_frn_history; 696 #endif 697 struct list_head i_lru; /* inode LRU list */ 698 struct list_head i_sb_list; 699 struct list_head i_wb_list; /* backing dev writeback list */ 700 union { 701 struct hlist_head i_dentry; 702 struct rcu_head i_rcu; 703 }; 704 atomic64_t i_version; 705 atomic64_t i_sequence; /* see futex */ 706 atomic_t i_count; 707 atomic_t i_dio_count; 708 atomic_t i_writecount; 709 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING) 710 atomic_t i_readcount; /* struct files open RO */ 711 #endif 712 union { 713 const struct file_operations *i_fop; /* former ->i_op->default_file_ops */ 714 void (*free_inode)(struct inode *); 715 }; 716 struct file_lock_context *i_flctx; 717 struct address_space i_data; 718 struct list_head i_devices; 719 union { 720 struct pipe_inode_info *i_pipe; 721 struct cdev *i_cdev; 722 char *i_link; 723 unsigned i_dir_seq; 724 }; 725 726 __u32 i_generation; 727 728 #ifdef CONFIG_FSNOTIFY 729 __u32 i_fsnotify_mask; /* all events this inode cares about */ 730 struct fsnotify_mark_connector __rcu *i_fsnotify_marks; 731 #endif 732 733 #ifdef CONFIG_FS_ENCRYPTION 734 struct fscrypt_inode_info *i_crypt_info; 735 #endif 736 737 #ifdef CONFIG_FS_VERITY 738 struct fsverity_info *i_verity_info; 739 #endif 740 741 void *i_private; /* fs or device private pointer */ 742 } __randomize_layout; 743 744 struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode); 745 746 static inline unsigned int i_blocksize(const struct inode *node) 747 { 748 return (1 << node->i_blkbits); 749 } 750 751 static inline int inode_unhashed(struct inode *inode) 752 { 753 return hlist_unhashed(&inode->i_hash); 754 } 755 756 /* 757 * __mark_inode_dirty expects inodes to be hashed. Since we don't 758 * want special inodes in the fileset inode space, we make them 759 * appear hashed, but do not put on any lists. hlist_del() 760 * will work fine and require no locking. 761 */ 762 static inline void inode_fake_hash(struct inode *inode) 763 { 764 hlist_add_fake(&inode->i_hash); 765 } 766 767 /* 768 * inode->i_mutex nesting subclasses for the lock validator: 769 * 770 * 0: the object of the current VFS operation 771 * 1: parent 772 * 2: child/target 773 * 3: xattr 774 * 4: second non-directory 775 * 5: second parent (when locking independent directories in rename) 776 * 777 * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two 778 * non-directories at once. 779 * 780 * The locking order between these classes is 781 * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory 782 */ 783 enum inode_i_mutex_lock_class 784 { 785 I_MUTEX_NORMAL, 786 I_MUTEX_PARENT, 787 I_MUTEX_CHILD, 788 I_MUTEX_XATTR, 789 I_MUTEX_NONDIR2, 790 I_MUTEX_PARENT2, 791 }; 792 793 static inline void inode_lock(struct inode *inode) 794 { 795 down_write(&inode->i_rwsem); 796 } 797 798 static inline void inode_unlock(struct inode *inode) 799 { 800 up_write(&inode->i_rwsem); 801 } 802 803 static inline void inode_lock_shared(struct inode *inode) 804 { 805 down_read(&inode->i_rwsem); 806 } 807 808 static inline void inode_unlock_shared(struct inode *inode) 809 { 810 up_read(&inode->i_rwsem); 811 } 812 813 static inline int inode_trylock(struct inode *inode) 814 { 815 return down_write_trylock(&inode->i_rwsem); 816 } 817 818 static inline int inode_trylock_shared(struct inode *inode) 819 { 820 return down_read_trylock(&inode->i_rwsem); 821 } 822 823 static inline int inode_is_locked(struct inode *inode) 824 { 825 return rwsem_is_locked(&inode->i_rwsem); 826 } 827 828 static inline void inode_lock_nested(struct inode *inode, unsigned subclass) 829 { 830 down_write_nested(&inode->i_rwsem, subclass); 831 } 832 833 static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass) 834 { 835 down_read_nested(&inode->i_rwsem, subclass); 836 } 837 838 static inline void filemap_invalidate_lock(struct address_space *mapping) 839 { 840 down_write(&mapping->invalidate_lock); 841 } 842 843 static inline void filemap_invalidate_unlock(struct address_space *mapping) 844 { 845 up_write(&mapping->invalidate_lock); 846 } 847 848 static inline void filemap_invalidate_lock_shared(struct address_space *mapping) 849 { 850 down_read(&mapping->invalidate_lock); 851 } 852 853 static inline int filemap_invalidate_trylock_shared( 854 struct address_space *mapping) 855 { 856 return down_read_trylock(&mapping->invalidate_lock); 857 } 858 859 static inline void filemap_invalidate_unlock_shared( 860 struct address_space *mapping) 861 { 862 up_read(&mapping->invalidate_lock); 863 } 864 865 void lock_two_nondirectories(struct inode *, struct inode*); 866 void unlock_two_nondirectories(struct inode *, struct inode*); 867 868 void filemap_invalidate_lock_two(struct address_space *mapping1, 869 struct address_space *mapping2); 870 void filemap_invalidate_unlock_two(struct address_space *mapping1, 871 struct address_space *mapping2); 872 873 874 /* 875 * NOTE: in a 32bit arch with a preemptable kernel and 876 * an UP compile the i_size_read/write must be atomic 877 * with respect to the local cpu (unlike with preempt disabled), 878 * but they don't need to be atomic with respect to other cpus like in 879 * true SMP (so they need either to either locally disable irq around 880 * the read or for example on x86 they can be still implemented as a 881 * cmpxchg8b without the need of the lock prefix). For SMP compiles 882 * and 64bit archs it makes no difference if preempt is enabled or not. 883 */ 884 static inline loff_t i_size_read(const struct inode *inode) 885 { 886 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 887 loff_t i_size; 888 unsigned int seq; 889 890 do { 891 seq = read_seqcount_begin(&inode->i_size_seqcount); 892 i_size = inode->i_size; 893 } while (read_seqcount_retry(&inode->i_size_seqcount, seq)); 894 return i_size; 895 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION) 896 loff_t i_size; 897 898 preempt_disable(); 899 i_size = inode->i_size; 900 preempt_enable(); 901 return i_size; 902 #else 903 /* Pairs with smp_store_release() in i_size_write() */ 904 return smp_load_acquire(&inode->i_size); 905 #endif 906 } 907 908 /* 909 * NOTE: unlike i_size_read(), i_size_write() does need locking around it 910 * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount 911 * can be lost, resulting in subsequent i_size_read() calls spinning forever. 912 */ 913 static inline void i_size_write(struct inode *inode, loff_t i_size) 914 { 915 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 916 preempt_disable(); 917 write_seqcount_begin(&inode->i_size_seqcount); 918 inode->i_size = i_size; 919 write_seqcount_end(&inode->i_size_seqcount); 920 preempt_enable(); 921 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION) 922 preempt_disable(); 923 inode->i_size = i_size; 924 preempt_enable(); 925 #else 926 /* 927 * Pairs with smp_load_acquire() in i_size_read() to ensure 928 * changes related to inode size (such as page contents) are 929 * visible before we see the changed inode size. 930 */ 931 smp_store_release(&inode->i_size, i_size); 932 #endif 933 } 934 935 static inline unsigned iminor(const struct inode *inode) 936 { 937 return MINOR(inode->i_rdev); 938 } 939 940 static inline unsigned imajor(const struct inode *inode) 941 { 942 return MAJOR(inode->i_rdev); 943 } 944 945 struct fown_struct { 946 rwlock_t lock; /* protects pid, uid, euid fields */ 947 struct pid *pid; /* pid or -pgrp where SIGIO should be sent */ 948 enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */ 949 kuid_t uid, euid; /* uid/euid of process setting the owner */ 950 int signum; /* posix.1b rt signal to be delivered on IO */ 951 }; 952 953 /** 954 * struct file_ra_state - Track a file's readahead state. 955 * @start: Where the most recent readahead started. 956 * @size: Number of pages read in the most recent readahead. 957 * @async_size: Numer of pages that were/are not needed immediately 958 * and so were/are genuinely "ahead". Start next readahead when 959 * the first of these pages is accessed. 960 * @ra_pages: Maximum size of a readahead request, copied from the bdi. 961 * @mmap_miss: How many mmap accesses missed in the page cache. 962 * @prev_pos: The last byte in the most recent read request. 963 * 964 * When this structure is passed to ->readahead(), the "most recent" 965 * readahead means the current readahead. 966 */ 967 struct file_ra_state { 968 pgoff_t start; 969 unsigned int size; 970 unsigned int async_size; 971 unsigned int ra_pages; 972 unsigned int mmap_miss; 973 loff_t prev_pos; 974 }; 975 976 /* 977 * Check if @index falls in the readahead windows. 978 */ 979 static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index) 980 { 981 return (index >= ra->start && 982 index < ra->start + ra->size); 983 } 984 985 /* 986 * f_{lock,count,pos_lock} members can be highly contended and share 987 * the same cacheline. f_{lock,mode} are very frequently used together 988 * and so share the same cacheline as well. The read-mostly 989 * f_{path,inode,op} are kept on a separate cacheline. 990 */ 991 struct file { 992 union { 993 /* fput() uses task work when closing and freeing file (default). */ 994 struct callback_head f_task_work; 995 /* fput() must use workqueue (most kernel threads). */ 996 struct llist_node f_llist; 997 unsigned int f_iocb_flags; 998 }; 999 1000 /* 1001 * Protects f_ep, f_flags. 1002 * Must not be taken from IRQ context. 1003 */ 1004 spinlock_t f_lock; 1005 fmode_t f_mode; 1006 atomic_long_t f_count; 1007 struct mutex f_pos_lock; 1008 loff_t f_pos; 1009 unsigned int f_flags; 1010 struct fown_struct f_owner; 1011 const struct cred *f_cred; 1012 struct file_ra_state f_ra; 1013 struct path f_path; 1014 struct inode *f_inode; /* cached value */ 1015 const struct file_operations *f_op; 1016 1017 u64 f_version; 1018 #ifdef CONFIG_SECURITY 1019 void *f_security; 1020 #endif 1021 /* needed for tty driver, and maybe others */ 1022 void *private_data; 1023 1024 #ifdef CONFIG_EPOLL 1025 /* Used by fs/eventpoll.c to link all the hooks to this file */ 1026 struct hlist_head *f_ep; 1027 #endif /* #ifdef CONFIG_EPOLL */ 1028 struct address_space *f_mapping; 1029 errseq_t f_wb_err; 1030 errseq_t f_sb_err; /* for syncfs */ 1031 } __randomize_layout 1032 __attribute__((aligned(4))); /* lest something weird decides that 2 is OK */ 1033 1034 struct file_handle { 1035 __u32 handle_bytes; 1036 int handle_type; 1037 /* file identifier */ 1038 unsigned char f_handle[]; 1039 }; 1040 1041 static inline struct file *get_file(struct file *f) 1042 { 1043 atomic_long_inc(&f->f_count); 1044 return f; 1045 } 1046 1047 struct file *get_file_rcu(struct file __rcu **f); 1048 struct file *get_file_active(struct file **f); 1049 1050 #define file_count(x) atomic_long_read(&(x)->f_count) 1051 1052 #define MAX_NON_LFS ((1UL<<31) - 1) 1053 1054 /* Page cache limit. The filesystems should put that into their s_maxbytes 1055 limits, otherwise bad things can happen in VM. */ 1056 #if BITS_PER_LONG==32 1057 #define MAX_LFS_FILESIZE ((loff_t)ULONG_MAX << PAGE_SHIFT) 1058 #elif BITS_PER_LONG==64 1059 #define MAX_LFS_FILESIZE ((loff_t)LLONG_MAX) 1060 #endif 1061 1062 /* legacy typedef, should eventually be removed */ 1063 typedef void *fl_owner_t; 1064 1065 struct file_lock; 1066 struct file_lease; 1067 1068 /* The following constant reflects the upper bound of the file/locking space */ 1069 #ifndef OFFSET_MAX 1070 #define OFFSET_MAX type_max(loff_t) 1071 #define OFFT_OFFSET_MAX type_max(off_t) 1072 #endif 1073 1074 extern void send_sigio(struct fown_struct *fown, int fd, int band); 1075 1076 static inline struct inode *file_inode(const struct file *f) 1077 { 1078 return f->f_inode; 1079 } 1080 1081 /* 1082 * file_dentry() is a relic from the days that overlayfs was using files with a 1083 * "fake" path, meaning, f_path on overlayfs and f_inode on underlying fs. 1084 * In those days, file_dentry() was needed to get the underlying fs dentry that 1085 * matches f_inode. 1086 * Files with "fake" path should not exist nowadays, so use an assertion to make 1087 * sure that file_dentry() was not papering over filesystem bugs. 1088 */ 1089 static inline struct dentry *file_dentry(const struct file *file) 1090 { 1091 struct dentry *dentry = file->f_path.dentry; 1092 1093 WARN_ON_ONCE(d_inode(dentry) != file_inode(file)); 1094 return dentry; 1095 } 1096 1097 struct fasync_struct { 1098 rwlock_t fa_lock; 1099 int magic; 1100 int fa_fd; 1101 struct fasync_struct *fa_next; /* singly linked list */ 1102 struct file *fa_file; 1103 struct rcu_head fa_rcu; 1104 }; 1105 1106 #define FASYNC_MAGIC 0x4601 1107 1108 /* SMP safe fasync helpers: */ 1109 extern int fasync_helper(int, struct file *, int, struct fasync_struct **); 1110 extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *); 1111 extern int fasync_remove_entry(struct file *, struct fasync_struct **); 1112 extern struct fasync_struct *fasync_alloc(void); 1113 extern void fasync_free(struct fasync_struct *); 1114 1115 /* can be called from interrupts */ 1116 extern void kill_fasync(struct fasync_struct **, int, int); 1117 1118 extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force); 1119 extern int f_setown(struct file *filp, int who, int force); 1120 extern void f_delown(struct file *filp); 1121 extern pid_t f_getown(struct file *filp); 1122 extern int send_sigurg(struct fown_struct *fown); 1123 1124 /* 1125 * sb->s_flags. Note that these mirror the equivalent MS_* flags where 1126 * represented in both. 1127 */ 1128 #define SB_RDONLY BIT(0) /* Mount read-only */ 1129 #define SB_NOSUID BIT(1) /* Ignore suid and sgid bits */ 1130 #define SB_NODEV BIT(2) /* Disallow access to device special files */ 1131 #define SB_NOEXEC BIT(3) /* Disallow program execution */ 1132 #define SB_SYNCHRONOUS BIT(4) /* Writes are synced at once */ 1133 #define SB_MANDLOCK BIT(6) /* Allow mandatory locks on an FS */ 1134 #define SB_DIRSYNC BIT(7) /* Directory modifications are synchronous */ 1135 #define SB_NOATIME BIT(10) /* Do not update access times. */ 1136 #define SB_NODIRATIME BIT(11) /* Do not update directory access times */ 1137 #define SB_SILENT BIT(15) 1138 #define SB_POSIXACL BIT(16) /* Supports POSIX ACLs */ 1139 #define SB_INLINECRYPT BIT(17) /* Use blk-crypto for encrypted files */ 1140 #define SB_KERNMOUNT BIT(22) /* this is a kern_mount call */ 1141 #define SB_I_VERSION BIT(23) /* Update inode I_version field */ 1142 #define SB_LAZYTIME BIT(25) /* Update the on-disk [acm]times lazily */ 1143 1144 /* These sb flags are internal to the kernel */ 1145 #define SB_DEAD BIT(21) 1146 #define SB_DYING BIT(24) 1147 #define SB_SUBMOUNT BIT(26) 1148 #define SB_FORCE BIT(27) 1149 #define SB_NOSEC BIT(28) 1150 #define SB_BORN BIT(29) 1151 #define SB_ACTIVE BIT(30) 1152 #define SB_NOUSER BIT(31) 1153 1154 /* These flags relate to encoding and casefolding */ 1155 #define SB_ENC_STRICT_MODE_FL (1 << 0) 1156 1157 #define sb_has_strict_encoding(sb) \ 1158 (sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL) 1159 1160 /* 1161 * Umount options 1162 */ 1163 1164 #define MNT_FORCE 0x00000001 /* Attempt to forcibily umount */ 1165 #define MNT_DETACH 0x00000002 /* Just detach from the tree */ 1166 #define MNT_EXPIRE 0x00000004 /* Mark for expiry */ 1167 #define UMOUNT_NOFOLLOW 0x00000008 /* Don't follow symlink on umount */ 1168 #define UMOUNT_UNUSED 0x80000000 /* Flag guaranteed to be unused */ 1169 1170 /* sb->s_iflags */ 1171 #define SB_I_CGROUPWB 0x00000001 /* cgroup-aware writeback enabled */ 1172 #define SB_I_NOEXEC 0x00000002 /* Ignore executables on this fs */ 1173 #define SB_I_NODEV 0x00000004 /* Ignore devices on this fs */ 1174 #define SB_I_STABLE_WRITES 0x00000008 /* don't modify blks until WB is done */ 1175 1176 /* sb->s_iflags to limit user namespace mounts */ 1177 #define SB_I_USERNS_VISIBLE 0x00000010 /* fstype already mounted */ 1178 #define SB_I_IMA_UNVERIFIABLE_SIGNATURE 0x00000020 1179 #define SB_I_UNTRUSTED_MOUNTER 0x00000040 1180 #define SB_I_EVM_UNSUPPORTED 0x00000080 1181 1182 #define SB_I_SKIP_SYNC 0x00000100 /* Skip superblock at global sync */ 1183 #define SB_I_PERSB_BDI 0x00000200 /* has a per-sb bdi */ 1184 #define SB_I_TS_EXPIRY_WARNED 0x00000400 /* warned about timestamp range expiry */ 1185 #define SB_I_RETIRED 0x00000800 /* superblock shouldn't be reused */ 1186 #define SB_I_NOUMASK 0x00001000 /* VFS does not apply umask */ 1187 1188 /* Possible states of 'frozen' field */ 1189 enum { 1190 SB_UNFROZEN = 0, /* FS is unfrozen */ 1191 SB_FREEZE_WRITE = 1, /* Writes, dir ops, ioctls frozen */ 1192 SB_FREEZE_PAGEFAULT = 2, /* Page faults stopped as well */ 1193 SB_FREEZE_FS = 3, /* For internal FS use (e.g. to stop 1194 * internal threads if needed) */ 1195 SB_FREEZE_COMPLETE = 4, /* ->freeze_fs finished successfully */ 1196 }; 1197 1198 #define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1) 1199 1200 struct sb_writers { 1201 unsigned short frozen; /* Is sb frozen? */ 1202 int freeze_kcount; /* How many kernel freeze requests? */ 1203 int freeze_ucount; /* How many userspace freeze requests? */ 1204 struct percpu_rw_semaphore rw_sem[SB_FREEZE_LEVELS]; 1205 }; 1206 1207 struct super_block { 1208 struct list_head s_list; /* Keep this first */ 1209 dev_t s_dev; /* search index; _not_ kdev_t */ 1210 unsigned char s_blocksize_bits; 1211 unsigned long s_blocksize; 1212 loff_t s_maxbytes; /* Max file size */ 1213 struct file_system_type *s_type; 1214 const struct super_operations *s_op; 1215 const struct dquot_operations *dq_op; 1216 const struct quotactl_ops *s_qcop; 1217 const struct export_operations *s_export_op; 1218 unsigned long s_flags; 1219 unsigned long s_iflags; /* internal SB_I_* flags */ 1220 unsigned long s_magic; 1221 struct dentry *s_root; 1222 struct rw_semaphore s_umount; 1223 int s_count; 1224 atomic_t s_active; 1225 #ifdef CONFIG_SECURITY 1226 void *s_security; 1227 #endif 1228 const struct xattr_handler * const *s_xattr; 1229 #ifdef CONFIG_FS_ENCRYPTION 1230 const struct fscrypt_operations *s_cop; 1231 struct fscrypt_keyring *s_master_keys; /* master crypto keys in use */ 1232 #endif 1233 #ifdef CONFIG_FS_VERITY 1234 const struct fsverity_operations *s_vop; 1235 #endif 1236 #if IS_ENABLED(CONFIG_UNICODE) 1237 struct unicode_map *s_encoding; 1238 __u16 s_encoding_flags; 1239 #endif 1240 struct hlist_bl_head s_roots; /* alternate root dentries for NFS */ 1241 struct list_head s_mounts; /* list of mounts; _not_ for fs use */ 1242 struct block_device *s_bdev; /* can go away once we use an accessor for @s_bdev_file */ 1243 struct file *s_bdev_file; 1244 struct backing_dev_info *s_bdi; 1245 struct mtd_info *s_mtd; 1246 struct hlist_node s_instances; 1247 unsigned int s_quota_types; /* Bitmask of supported quota types */ 1248 struct quota_info s_dquot; /* Diskquota specific options */ 1249 1250 struct sb_writers s_writers; 1251 1252 /* 1253 * Keep s_fs_info, s_time_gran, s_fsnotify_mask, and 1254 * s_fsnotify_marks together for cache efficiency. They are frequently 1255 * accessed and rarely modified. 1256 */ 1257 void *s_fs_info; /* Filesystem private info */ 1258 1259 /* Granularity of c/m/atime in ns (cannot be worse than a second) */ 1260 u32 s_time_gran; 1261 /* Time limits for c/m/atime in seconds */ 1262 time64_t s_time_min; 1263 time64_t s_time_max; 1264 #ifdef CONFIG_FSNOTIFY 1265 __u32 s_fsnotify_mask; 1266 struct fsnotify_mark_connector __rcu *s_fsnotify_marks; 1267 #endif 1268 1269 /* 1270 * q: why are s_id and s_sysfs_name not the same? both are human 1271 * readable strings that identify the filesystem 1272 * a: s_id is allowed to change at runtime; it's used in log messages, 1273 * and we want to when a device starts out as single device (s_id is dev 1274 * name) but then a device is hot added and we have to switch to 1275 * identifying it by UUID 1276 * but s_sysfs_name is a handle for programmatic access, and can't 1277 * change at runtime 1278 */ 1279 char s_id[32]; /* Informational name */ 1280 uuid_t s_uuid; /* UUID */ 1281 u8 s_uuid_len; /* Default 16, possibly smaller for weird filesystems */ 1282 1283 /* if set, fs shows up under sysfs at /sys/fs/$FSTYP/s_sysfs_name */ 1284 char s_sysfs_name[UUID_STRING_LEN + 1]; 1285 1286 unsigned int s_max_links; 1287 1288 /* 1289 * The next field is for VFS *only*. No filesystems have any business 1290 * even looking at it. You had been warned. 1291 */ 1292 struct mutex s_vfs_rename_mutex; /* Kludge */ 1293 1294 /* 1295 * Filesystem subtype. If non-empty the filesystem type field 1296 * in /proc/mounts will be "type.subtype" 1297 */ 1298 const char *s_subtype; 1299 1300 const struct dentry_operations *s_d_op; /* default d_op for dentries */ 1301 1302 struct shrinker *s_shrink; /* per-sb shrinker handle */ 1303 1304 /* Number of inodes with nlink == 0 but still referenced */ 1305 atomic_long_t s_remove_count; 1306 1307 /* 1308 * Number of inode/mount/sb objects that are being watched, note that 1309 * inodes objects are currently double-accounted. 1310 */ 1311 atomic_long_t s_fsnotify_connectors; 1312 1313 /* Read-only state of the superblock is being changed */ 1314 int s_readonly_remount; 1315 1316 /* per-sb errseq_t for reporting writeback errors via syncfs */ 1317 errseq_t s_wb_err; 1318 1319 /* AIO completions deferred from interrupt context */ 1320 struct workqueue_struct *s_dio_done_wq; 1321 struct hlist_head s_pins; 1322 1323 /* 1324 * Owning user namespace and default context in which to 1325 * interpret filesystem uids, gids, quotas, device nodes, 1326 * xattrs and security labels. 1327 */ 1328 struct user_namespace *s_user_ns; 1329 1330 /* 1331 * The list_lru structure is essentially just a pointer to a table 1332 * of per-node lru lists, each of which has its own spinlock. 1333 * There is no need to put them into separate cachelines. 1334 */ 1335 struct list_lru s_dentry_lru; 1336 struct list_lru s_inode_lru; 1337 struct rcu_head rcu; 1338 struct work_struct destroy_work; 1339 1340 struct mutex s_sync_lock; /* sync serialisation lock */ 1341 1342 /* 1343 * Indicates how deep in a filesystem stack this SB is 1344 */ 1345 int s_stack_depth; 1346 1347 /* s_inode_list_lock protects s_inodes */ 1348 spinlock_t s_inode_list_lock ____cacheline_aligned_in_smp; 1349 struct list_head s_inodes; /* all inodes */ 1350 1351 spinlock_t s_inode_wblist_lock; 1352 struct list_head s_inodes_wb; /* writeback inodes */ 1353 } __randomize_layout; 1354 1355 static inline struct user_namespace *i_user_ns(const struct inode *inode) 1356 { 1357 return inode->i_sb->s_user_ns; 1358 } 1359 1360 /* Helper functions so that in most cases filesystems will 1361 * not need to deal directly with kuid_t and kgid_t and can 1362 * instead deal with the raw numeric values that are stored 1363 * in the filesystem. 1364 */ 1365 static inline uid_t i_uid_read(const struct inode *inode) 1366 { 1367 return from_kuid(i_user_ns(inode), inode->i_uid); 1368 } 1369 1370 static inline gid_t i_gid_read(const struct inode *inode) 1371 { 1372 return from_kgid(i_user_ns(inode), inode->i_gid); 1373 } 1374 1375 static inline void i_uid_write(struct inode *inode, uid_t uid) 1376 { 1377 inode->i_uid = make_kuid(i_user_ns(inode), uid); 1378 } 1379 1380 static inline void i_gid_write(struct inode *inode, gid_t gid) 1381 { 1382 inode->i_gid = make_kgid(i_user_ns(inode), gid); 1383 } 1384 1385 /** 1386 * i_uid_into_vfsuid - map an inode's i_uid down according to an idmapping 1387 * @idmap: idmap of the mount the inode was found from 1388 * @inode: inode to map 1389 * 1390 * Return: whe inode's i_uid mapped down according to @idmap. 1391 * If the inode's i_uid has no mapping INVALID_VFSUID is returned. 1392 */ 1393 static inline vfsuid_t i_uid_into_vfsuid(struct mnt_idmap *idmap, 1394 const struct inode *inode) 1395 { 1396 return make_vfsuid(idmap, i_user_ns(inode), inode->i_uid); 1397 } 1398 1399 /** 1400 * i_uid_needs_update - check whether inode's i_uid needs to be updated 1401 * @idmap: idmap of the mount the inode was found from 1402 * @attr: the new attributes of @inode 1403 * @inode: the inode to update 1404 * 1405 * Check whether the $inode's i_uid field needs to be updated taking idmapped 1406 * mounts into account if the filesystem supports it. 1407 * 1408 * Return: true if @inode's i_uid field needs to be updated, false if not. 1409 */ 1410 static inline bool i_uid_needs_update(struct mnt_idmap *idmap, 1411 const struct iattr *attr, 1412 const struct inode *inode) 1413 { 1414 return ((attr->ia_valid & ATTR_UID) && 1415 !vfsuid_eq(attr->ia_vfsuid, 1416 i_uid_into_vfsuid(idmap, inode))); 1417 } 1418 1419 /** 1420 * i_uid_update - update @inode's i_uid field 1421 * @idmap: idmap of the mount the inode was found from 1422 * @attr: the new attributes of @inode 1423 * @inode: the inode to update 1424 * 1425 * Safely update @inode's i_uid field translating the vfsuid of any idmapped 1426 * mount into the filesystem kuid. 1427 */ 1428 static inline void i_uid_update(struct mnt_idmap *idmap, 1429 const struct iattr *attr, 1430 struct inode *inode) 1431 { 1432 if (attr->ia_valid & ATTR_UID) 1433 inode->i_uid = from_vfsuid(idmap, i_user_ns(inode), 1434 attr->ia_vfsuid); 1435 } 1436 1437 /** 1438 * i_gid_into_vfsgid - map an inode's i_gid down according to an idmapping 1439 * @idmap: idmap of the mount the inode was found from 1440 * @inode: inode to map 1441 * 1442 * Return: the inode's i_gid mapped down according to @idmap. 1443 * If the inode's i_gid has no mapping INVALID_VFSGID is returned. 1444 */ 1445 static inline vfsgid_t i_gid_into_vfsgid(struct mnt_idmap *idmap, 1446 const struct inode *inode) 1447 { 1448 return make_vfsgid(idmap, i_user_ns(inode), inode->i_gid); 1449 } 1450 1451 /** 1452 * i_gid_needs_update - check whether inode's i_gid needs to be updated 1453 * @idmap: idmap of the mount the inode was found from 1454 * @attr: the new attributes of @inode 1455 * @inode: the inode to update 1456 * 1457 * Check whether the $inode's i_gid field needs to be updated taking idmapped 1458 * mounts into account if the filesystem supports it. 1459 * 1460 * Return: true if @inode's i_gid field needs to be updated, false if not. 1461 */ 1462 static inline bool i_gid_needs_update(struct mnt_idmap *idmap, 1463 const struct iattr *attr, 1464 const struct inode *inode) 1465 { 1466 return ((attr->ia_valid & ATTR_GID) && 1467 !vfsgid_eq(attr->ia_vfsgid, 1468 i_gid_into_vfsgid(idmap, inode))); 1469 } 1470 1471 /** 1472 * i_gid_update - update @inode's i_gid field 1473 * @idmap: idmap of the mount the inode was found from 1474 * @attr: the new attributes of @inode 1475 * @inode: the inode to update 1476 * 1477 * Safely update @inode's i_gid field translating the vfsgid of any idmapped 1478 * mount into the filesystem kgid. 1479 */ 1480 static inline void i_gid_update(struct mnt_idmap *idmap, 1481 const struct iattr *attr, 1482 struct inode *inode) 1483 { 1484 if (attr->ia_valid & ATTR_GID) 1485 inode->i_gid = from_vfsgid(idmap, i_user_ns(inode), 1486 attr->ia_vfsgid); 1487 } 1488 1489 /** 1490 * inode_fsuid_set - initialize inode's i_uid field with callers fsuid 1491 * @inode: inode to initialize 1492 * @idmap: idmap of the mount the inode was found from 1493 * 1494 * Initialize the i_uid field of @inode. If the inode was found/created via 1495 * an idmapped mount map the caller's fsuid according to @idmap. 1496 */ 1497 static inline void inode_fsuid_set(struct inode *inode, 1498 struct mnt_idmap *idmap) 1499 { 1500 inode->i_uid = mapped_fsuid(idmap, i_user_ns(inode)); 1501 } 1502 1503 /** 1504 * inode_fsgid_set - initialize inode's i_gid field with callers fsgid 1505 * @inode: inode to initialize 1506 * @idmap: idmap of the mount the inode was found from 1507 * 1508 * Initialize the i_gid field of @inode. If the inode was found/created via 1509 * an idmapped mount map the caller's fsgid according to @idmap. 1510 */ 1511 static inline void inode_fsgid_set(struct inode *inode, 1512 struct mnt_idmap *idmap) 1513 { 1514 inode->i_gid = mapped_fsgid(idmap, i_user_ns(inode)); 1515 } 1516 1517 /** 1518 * fsuidgid_has_mapping() - check whether caller's fsuid/fsgid is mapped 1519 * @sb: the superblock we want a mapping in 1520 * @idmap: idmap of the relevant mount 1521 * 1522 * Check whether the caller's fsuid and fsgid have a valid mapping in the 1523 * s_user_ns of the superblock @sb. If the caller is on an idmapped mount map 1524 * the caller's fsuid and fsgid according to the @idmap first. 1525 * 1526 * Return: true if fsuid and fsgid is mapped, false if not. 1527 */ 1528 static inline bool fsuidgid_has_mapping(struct super_block *sb, 1529 struct mnt_idmap *idmap) 1530 { 1531 struct user_namespace *fs_userns = sb->s_user_ns; 1532 kuid_t kuid; 1533 kgid_t kgid; 1534 1535 kuid = mapped_fsuid(idmap, fs_userns); 1536 if (!uid_valid(kuid)) 1537 return false; 1538 kgid = mapped_fsgid(idmap, fs_userns); 1539 if (!gid_valid(kgid)) 1540 return false; 1541 return kuid_has_mapping(fs_userns, kuid) && 1542 kgid_has_mapping(fs_userns, kgid); 1543 } 1544 1545 struct timespec64 current_time(struct inode *inode); 1546 struct timespec64 inode_set_ctime_current(struct inode *inode); 1547 1548 static inline time64_t inode_get_atime_sec(const struct inode *inode) 1549 { 1550 return inode->__i_atime.tv_sec; 1551 } 1552 1553 static inline long inode_get_atime_nsec(const struct inode *inode) 1554 { 1555 return inode->__i_atime.tv_nsec; 1556 } 1557 1558 static inline struct timespec64 inode_get_atime(const struct inode *inode) 1559 { 1560 return inode->__i_atime; 1561 } 1562 1563 static inline struct timespec64 inode_set_atime_to_ts(struct inode *inode, 1564 struct timespec64 ts) 1565 { 1566 inode->__i_atime = ts; 1567 return ts; 1568 } 1569 1570 static inline struct timespec64 inode_set_atime(struct inode *inode, 1571 time64_t sec, long nsec) 1572 { 1573 struct timespec64 ts = { .tv_sec = sec, 1574 .tv_nsec = nsec }; 1575 return inode_set_atime_to_ts(inode, ts); 1576 } 1577 1578 static inline time64_t inode_get_mtime_sec(const struct inode *inode) 1579 { 1580 return inode->__i_mtime.tv_sec; 1581 } 1582 1583 static inline long inode_get_mtime_nsec(const struct inode *inode) 1584 { 1585 return inode->__i_mtime.tv_nsec; 1586 } 1587 1588 static inline struct timespec64 inode_get_mtime(const struct inode *inode) 1589 { 1590 return inode->__i_mtime; 1591 } 1592 1593 static inline struct timespec64 inode_set_mtime_to_ts(struct inode *inode, 1594 struct timespec64 ts) 1595 { 1596 inode->__i_mtime = ts; 1597 return ts; 1598 } 1599 1600 static inline struct timespec64 inode_set_mtime(struct inode *inode, 1601 time64_t sec, long nsec) 1602 { 1603 struct timespec64 ts = { .tv_sec = sec, 1604 .tv_nsec = nsec }; 1605 return inode_set_mtime_to_ts(inode, ts); 1606 } 1607 1608 static inline time64_t inode_get_ctime_sec(const struct inode *inode) 1609 { 1610 return inode->__i_ctime.tv_sec; 1611 } 1612 1613 static inline long inode_get_ctime_nsec(const struct inode *inode) 1614 { 1615 return inode->__i_ctime.tv_nsec; 1616 } 1617 1618 static inline struct timespec64 inode_get_ctime(const struct inode *inode) 1619 { 1620 return inode->__i_ctime; 1621 } 1622 1623 static inline struct timespec64 inode_set_ctime_to_ts(struct inode *inode, 1624 struct timespec64 ts) 1625 { 1626 inode->__i_ctime = ts; 1627 return ts; 1628 } 1629 1630 /** 1631 * inode_set_ctime - set the ctime in the inode 1632 * @inode: inode in which to set the ctime 1633 * @sec: tv_sec value to set 1634 * @nsec: tv_nsec value to set 1635 * 1636 * Set the ctime in @inode to { @sec, @nsec } 1637 */ 1638 static inline struct timespec64 inode_set_ctime(struct inode *inode, 1639 time64_t sec, long nsec) 1640 { 1641 struct timespec64 ts = { .tv_sec = sec, 1642 .tv_nsec = nsec }; 1643 1644 return inode_set_ctime_to_ts(inode, ts); 1645 } 1646 1647 struct timespec64 simple_inode_init_ts(struct inode *inode); 1648 1649 /* 1650 * Snapshotting support. 1651 */ 1652 1653 /* 1654 * These are internal functions, please use sb_start_{write,pagefault,intwrite} 1655 * instead. 1656 */ 1657 static inline void __sb_end_write(struct super_block *sb, int level) 1658 { 1659 percpu_up_read(sb->s_writers.rw_sem + level-1); 1660 } 1661 1662 static inline void __sb_start_write(struct super_block *sb, int level) 1663 { 1664 percpu_down_read(sb->s_writers.rw_sem + level - 1); 1665 } 1666 1667 static inline bool __sb_start_write_trylock(struct super_block *sb, int level) 1668 { 1669 return percpu_down_read_trylock(sb->s_writers.rw_sem + level - 1); 1670 } 1671 1672 #define __sb_writers_acquired(sb, lev) \ 1673 percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_) 1674 #define __sb_writers_release(sb, lev) \ 1675 percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_) 1676 1677 /** 1678 * __sb_write_started - check if sb freeze level is held 1679 * @sb: the super we write to 1680 * @level: the freeze level 1681 * 1682 * * > 0 - sb freeze level is held 1683 * * 0 - sb freeze level is not held 1684 * * < 0 - !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN 1685 */ 1686 static inline int __sb_write_started(const struct super_block *sb, int level) 1687 { 1688 return lockdep_is_held_type(sb->s_writers.rw_sem + level - 1, 1); 1689 } 1690 1691 /** 1692 * sb_write_started - check if SB_FREEZE_WRITE is held 1693 * @sb: the super we write to 1694 * 1695 * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN. 1696 */ 1697 static inline bool sb_write_started(const struct super_block *sb) 1698 { 1699 return __sb_write_started(sb, SB_FREEZE_WRITE); 1700 } 1701 1702 /** 1703 * sb_write_not_started - check if SB_FREEZE_WRITE is not held 1704 * @sb: the super we write to 1705 * 1706 * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN. 1707 */ 1708 static inline bool sb_write_not_started(const struct super_block *sb) 1709 { 1710 return __sb_write_started(sb, SB_FREEZE_WRITE) <= 0; 1711 } 1712 1713 /** 1714 * file_write_started - check if SB_FREEZE_WRITE is held 1715 * @file: the file we write to 1716 * 1717 * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN. 1718 * May be false positive with !S_ISREG, because file_start_write() has 1719 * no effect on !S_ISREG. 1720 */ 1721 static inline bool file_write_started(const struct file *file) 1722 { 1723 if (!S_ISREG(file_inode(file)->i_mode)) 1724 return true; 1725 return sb_write_started(file_inode(file)->i_sb); 1726 } 1727 1728 /** 1729 * file_write_not_started - check if SB_FREEZE_WRITE is not held 1730 * @file: the file we write to 1731 * 1732 * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN. 1733 * May be false positive with !S_ISREG, because file_start_write() has 1734 * no effect on !S_ISREG. 1735 */ 1736 static inline bool file_write_not_started(const struct file *file) 1737 { 1738 if (!S_ISREG(file_inode(file)->i_mode)) 1739 return true; 1740 return sb_write_not_started(file_inode(file)->i_sb); 1741 } 1742 1743 /** 1744 * sb_end_write - drop write access to a superblock 1745 * @sb: the super we wrote to 1746 * 1747 * Decrement number of writers to the filesystem. Wake up possible waiters 1748 * wanting to freeze the filesystem. 1749 */ 1750 static inline void sb_end_write(struct super_block *sb) 1751 { 1752 __sb_end_write(sb, SB_FREEZE_WRITE); 1753 } 1754 1755 /** 1756 * sb_end_pagefault - drop write access to a superblock from a page fault 1757 * @sb: the super we wrote to 1758 * 1759 * Decrement number of processes handling write page fault to the filesystem. 1760 * Wake up possible waiters wanting to freeze the filesystem. 1761 */ 1762 static inline void sb_end_pagefault(struct super_block *sb) 1763 { 1764 __sb_end_write(sb, SB_FREEZE_PAGEFAULT); 1765 } 1766 1767 /** 1768 * sb_end_intwrite - drop write access to a superblock for internal fs purposes 1769 * @sb: the super we wrote to 1770 * 1771 * Decrement fs-internal number of writers to the filesystem. Wake up possible 1772 * waiters wanting to freeze the filesystem. 1773 */ 1774 static inline void sb_end_intwrite(struct super_block *sb) 1775 { 1776 __sb_end_write(sb, SB_FREEZE_FS); 1777 } 1778 1779 /** 1780 * sb_start_write - get write access to a superblock 1781 * @sb: the super we write to 1782 * 1783 * When a process wants to write data or metadata to a file system (i.e. dirty 1784 * a page or an inode), it should embed the operation in a sb_start_write() - 1785 * sb_end_write() pair to get exclusion against file system freezing. This 1786 * function increments number of writers preventing freezing. If the file 1787 * system is already frozen, the function waits until the file system is 1788 * thawed. 1789 * 1790 * Since freeze protection behaves as a lock, users have to preserve 1791 * ordering of freeze protection and other filesystem locks. Generally, 1792 * freeze protection should be the outermost lock. In particular, we have: 1793 * 1794 * sb_start_write 1795 * -> i_mutex (write path, truncate, directory ops, ...) 1796 * -> s_umount (freeze_super, thaw_super) 1797 */ 1798 static inline void sb_start_write(struct super_block *sb) 1799 { 1800 __sb_start_write(sb, SB_FREEZE_WRITE); 1801 } 1802 1803 static inline bool sb_start_write_trylock(struct super_block *sb) 1804 { 1805 return __sb_start_write_trylock(sb, SB_FREEZE_WRITE); 1806 } 1807 1808 /** 1809 * sb_start_pagefault - get write access to a superblock from a page fault 1810 * @sb: the super we write to 1811 * 1812 * When a process starts handling write page fault, it should embed the 1813 * operation into sb_start_pagefault() - sb_end_pagefault() pair to get 1814 * exclusion against file system freezing. This is needed since the page fault 1815 * is going to dirty a page. This function increments number of running page 1816 * faults preventing freezing. If the file system is already frozen, the 1817 * function waits until the file system is thawed. 1818 * 1819 * Since page fault freeze protection behaves as a lock, users have to preserve 1820 * ordering of freeze protection and other filesystem locks. It is advised to 1821 * put sb_start_pagefault() close to mmap_lock in lock ordering. Page fault 1822 * handling code implies lock dependency: 1823 * 1824 * mmap_lock 1825 * -> sb_start_pagefault 1826 */ 1827 static inline void sb_start_pagefault(struct super_block *sb) 1828 { 1829 __sb_start_write(sb, SB_FREEZE_PAGEFAULT); 1830 } 1831 1832 /** 1833 * sb_start_intwrite - get write access to a superblock for internal fs purposes 1834 * @sb: the super we write to 1835 * 1836 * This is the third level of protection against filesystem freezing. It is 1837 * free for use by a filesystem. The only requirement is that it must rank 1838 * below sb_start_pagefault. 1839 * 1840 * For example filesystem can call sb_start_intwrite() when starting a 1841 * transaction which somewhat eases handling of freezing for internal sources 1842 * of filesystem changes (internal fs threads, discarding preallocation on file 1843 * close, etc.). 1844 */ 1845 static inline void sb_start_intwrite(struct super_block *sb) 1846 { 1847 __sb_start_write(sb, SB_FREEZE_FS); 1848 } 1849 1850 static inline bool sb_start_intwrite_trylock(struct super_block *sb) 1851 { 1852 return __sb_start_write_trylock(sb, SB_FREEZE_FS); 1853 } 1854 1855 bool inode_owner_or_capable(struct mnt_idmap *idmap, 1856 const struct inode *inode); 1857 1858 /* 1859 * VFS helper functions.. 1860 */ 1861 int vfs_create(struct mnt_idmap *, struct inode *, 1862 struct dentry *, umode_t, bool); 1863 int vfs_mkdir(struct mnt_idmap *, struct inode *, 1864 struct dentry *, umode_t); 1865 int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *, 1866 umode_t, dev_t); 1867 int vfs_symlink(struct mnt_idmap *, struct inode *, 1868 struct dentry *, const char *); 1869 int vfs_link(struct dentry *, struct mnt_idmap *, struct inode *, 1870 struct dentry *, struct inode **); 1871 int vfs_rmdir(struct mnt_idmap *, struct inode *, struct dentry *); 1872 int vfs_unlink(struct mnt_idmap *, struct inode *, struct dentry *, 1873 struct inode **); 1874 1875 /** 1876 * struct renamedata - contains all information required for renaming 1877 * @old_mnt_idmap: idmap of the old mount the inode was found from 1878 * @old_dir: parent of source 1879 * @old_dentry: source 1880 * @new_mnt_idmap: idmap of the new mount the inode was found from 1881 * @new_dir: parent of destination 1882 * @new_dentry: destination 1883 * @delegated_inode: returns an inode needing a delegation break 1884 * @flags: rename flags 1885 */ 1886 struct renamedata { 1887 struct mnt_idmap *old_mnt_idmap; 1888 struct inode *old_dir; 1889 struct dentry *old_dentry; 1890 struct mnt_idmap *new_mnt_idmap; 1891 struct inode *new_dir; 1892 struct dentry *new_dentry; 1893 struct inode **delegated_inode; 1894 unsigned int flags; 1895 } __randomize_layout; 1896 1897 int vfs_rename(struct renamedata *); 1898 1899 static inline int vfs_whiteout(struct mnt_idmap *idmap, 1900 struct inode *dir, struct dentry *dentry) 1901 { 1902 return vfs_mknod(idmap, dir, dentry, S_IFCHR | WHITEOUT_MODE, 1903 WHITEOUT_DEV); 1904 } 1905 1906 struct file *kernel_tmpfile_open(struct mnt_idmap *idmap, 1907 const struct path *parentpath, 1908 umode_t mode, int open_flag, 1909 const struct cred *cred); 1910 struct file *kernel_file_open(const struct path *path, int flags, 1911 struct inode *inode, const struct cred *cred); 1912 1913 int vfs_mkobj(struct dentry *, umode_t, 1914 int (*f)(struct dentry *, umode_t, void *), 1915 void *); 1916 1917 int vfs_fchown(struct file *file, uid_t user, gid_t group); 1918 int vfs_fchmod(struct file *file, umode_t mode); 1919 int vfs_utimes(const struct path *path, struct timespec64 *times); 1920 1921 extern long vfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 1922 1923 #ifdef CONFIG_COMPAT 1924 extern long compat_ptr_ioctl(struct file *file, unsigned int cmd, 1925 unsigned long arg); 1926 #else 1927 #define compat_ptr_ioctl NULL 1928 #endif 1929 1930 /* 1931 * VFS file helper functions. 1932 */ 1933 void inode_init_owner(struct mnt_idmap *idmap, struct inode *inode, 1934 const struct inode *dir, umode_t mode); 1935 extern bool may_open_dev(const struct path *path); 1936 umode_t mode_strip_sgid(struct mnt_idmap *idmap, 1937 const struct inode *dir, umode_t mode); 1938 1939 /* 1940 * This is the "filldir" function type, used by readdir() to let 1941 * the kernel specify what kind of dirent layout it wants to have. 1942 * This allows the kernel to read directories into kernel space or 1943 * to have different dirent layouts depending on the binary type. 1944 * Return 'true' to keep going and 'false' if there are no more entries. 1945 */ 1946 struct dir_context; 1947 typedef bool (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64, 1948 unsigned); 1949 1950 struct dir_context { 1951 filldir_t actor; 1952 loff_t pos; 1953 }; 1954 1955 /* 1956 * These flags let !MMU mmap() govern direct device mapping vs immediate 1957 * copying more easily for MAP_PRIVATE, especially for ROM filesystems. 1958 * 1959 * NOMMU_MAP_COPY: Copy can be mapped (MAP_PRIVATE) 1960 * NOMMU_MAP_DIRECT: Can be mapped directly (MAP_SHARED) 1961 * NOMMU_MAP_READ: Can be mapped for reading 1962 * NOMMU_MAP_WRITE: Can be mapped for writing 1963 * NOMMU_MAP_EXEC: Can be mapped for execution 1964 */ 1965 #define NOMMU_MAP_COPY 0x00000001 1966 #define NOMMU_MAP_DIRECT 0x00000008 1967 #define NOMMU_MAP_READ VM_MAYREAD 1968 #define NOMMU_MAP_WRITE VM_MAYWRITE 1969 #define NOMMU_MAP_EXEC VM_MAYEXEC 1970 1971 #define NOMMU_VMFLAGS \ 1972 (NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC) 1973 1974 /* 1975 * These flags control the behavior of the remap_file_range function pointer. 1976 * If it is called with len == 0 that means "remap to end of source file". 1977 * See Documentation/filesystems/vfs.rst for more details about this call. 1978 * 1979 * REMAP_FILE_DEDUP: only remap if contents identical (i.e. deduplicate) 1980 * REMAP_FILE_CAN_SHORTEN: caller can handle a shortened request 1981 */ 1982 #define REMAP_FILE_DEDUP (1 << 0) 1983 #define REMAP_FILE_CAN_SHORTEN (1 << 1) 1984 1985 /* 1986 * These flags signal that the caller is ok with altering various aspects of 1987 * the behavior of the remap operation. The changes must be made by the 1988 * implementation; the vfs remap helper functions can take advantage of them. 1989 * Flags in this category exist to preserve the quirky behavior of the hoisted 1990 * btrfs clone/dedupe ioctls. 1991 */ 1992 #define REMAP_FILE_ADVISORY (REMAP_FILE_CAN_SHORTEN) 1993 1994 /* 1995 * These flags control the behavior of vfs_copy_file_range(). 1996 * They are not available to the user via syscall. 1997 * 1998 * COPY_FILE_SPLICE: call splice direct instead of fs clone/copy ops 1999 */ 2000 #define COPY_FILE_SPLICE (1 << 0) 2001 2002 struct iov_iter; 2003 struct io_uring_cmd; 2004 struct offset_ctx; 2005 2006 struct file_operations { 2007 struct module *owner; 2008 loff_t (*llseek) (struct file *, loff_t, int); 2009 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *); 2010 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *); 2011 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *); 2012 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *); 2013 int (*iopoll)(struct kiocb *kiocb, struct io_comp_batch *, 2014 unsigned int flags); 2015 int (*iterate_shared) (struct file *, struct dir_context *); 2016 __poll_t (*poll) (struct file *, struct poll_table_struct *); 2017 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long); 2018 long (*compat_ioctl) (struct file *, unsigned int, unsigned long); 2019 int (*mmap) (struct file *, struct vm_area_struct *); 2020 unsigned long mmap_supported_flags; 2021 int (*open) (struct inode *, struct file *); 2022 int (*flush) (struct file *, fl_owner_t id); 2023 int (*release) (struct inode *, struct file *); 2024 int (*fsync) (struct file *, loff_t, loff_t, int datasync); 2025 int (*fasync) (int, struct file *, int); 2026 int (*lock) (struct file *, int, struct file_lock *); 2027 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 2028 int (*check_flags)(int); 2029 int (*flock) (struct file *, int, struct file_lock *); 2030 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int); 2031 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int); 2032 void (*splice_eof)(struct file *file); 2033 int (*setlease)(struct file *, int, struct file_lease **, void **); 2034 long (*fallocate)(struct file *file, int mode, loff_t offset, 2035 loff_t len); 2036 void (*show_fdinfo)(struct seq_file *m, struct file *f); 2037 #ifndef CONFIG_MMU 2038 unsigned (*mmap_capabilities)(struct file *); 2039 #endif 2040 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *, 2041 loff_t, size_t, unsigned int); 2042 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in, 2043 struct file *file_out, loff_t pos_out, 2044 loff_t len, unsigned int remap_flags); 2045 int (*fadvise)(struct file *, loff_t, loff_t, int); 2046 int (*uring_cmd)(struct io_uring_cmd *ioucmd, unsigned int issue_flags); 2047 int (*uring_cmd_iopoll)(struct io_uring_cmd *, struct io_comp_batch *, 2048 unsigned int poll_flags); 2049 } __randomize_layout; 2050 2051 /* Wrap a directory iterator that needs exclusive inode access */ 2052 int wrap_directory_iterator(struct file *, struct dir_context *, 2053 int (*) (struct file *, struct dir_context *)); 2054 #define WRAP_DIR_ITER(x) \ 2055 static int shared_##x(struct file *file , struct dir_context *ctx) \ 2056 { return wrap_directory_iterator(file, ctx, x); } 2057 2058 struct inode_operations { 2059 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int); 2060 const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *); 2061 int (*permission) (struct mnt_idmap *, struct inode *, int); 2062 struct posix_acl * (*get_inode_acl)(struct inode *, int, bool); 2063 2064 int (*readlink) (struct dentry *, char __user *,int); 2065 2066 int (*create) (struct mnt_idmap *, struct inode *,struct dentry *, 2067 umode_t, bool); 2068 int (*link) (struct dentry *,struct inode *,struct dentry *); 2069 int (*unlink) (struct inode *,struct dentry *); 2070 int (*symlink) (struct mnt_idmap *, struct inode *,struct dentry *, 2071 const char *); 2072 int (*mkdir) (struct mnt_idmap *, struct inode *,struct dentry *, 2073 umode_t); 2074 int (*rmdir) (struct inode *,struct dentry *); 2075 int (*mknod) (struct mnt_idmap *, struct inode *,struct dentry *, 2076 umode_t,dev_t); 2077 int (*rename) (struct mnt_idmap *, struct inode *, struct dentry *, 2078 struct inode *, struct dentry *, unsigned int); 2079 int (*setattr) (struct mnt_idmap *, struct dentry *, struct iattr *); 2080 int (*getattr) (struct mnt_idmap *, const struct path *, 2081 struct kstat *, u32, unsigned int); 2082 ssize_t (*listxattr) (struct dentry *, char *, size_t); 2083 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start, 2084 u64 len); 2085 int (*update_time)(struct inode *, int); 2086 int (*atomic_open)(struct inode *, struct dentry *, 2087 struct file *, unsigned open_flag, 2088 umode_t create_mode); 2089 int (*tmpfile) (struct mnt_idmap *, struct inode *, 2090 struct file *, umode_t); 2091 struct posix_acl *(*get_acl)(struct mnt_idmap *, struct dentry *, 2092 int); 2093 int (*set_acl)(struct mnt_idmap *, struct dentry *, 2094 struct posix_acl *, int); 2095 int (*fileattr_set)(struct mnt_idmap *idmap, 2096 struct dentry *dentry, struct fileattr *fa); 2097 int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa); 2098 struct offset_ctx *(*get_offset_ctx)(struct inode *inode); 2099 } ____cacheline_aligned; 2100 2101 static inline ssize_t call_read_iter(struct file *file, struct kiocb *kio, 2102 struct iov_iter *iter) 2103 { 2104 return file->f_op->read_iter(kio, iter); 2105 } 2106 2107 static inline ssize_t call_write_iter(struct file *file, struct kiocb *kio, 2108 struct iov_iter *iter) 2109 { 2110 return file->f_op->write_iter(kio, iter); 2111 } 2112 2113 static inline int call_mmap(struct file *file, struct vm_area_struct *vma) 2114 { 2115 return file->f_op->mmap(file, vma); 2116 } 2117 2118 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *); 2119 extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *); 2120 extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *, 2121 loff_t, size_t, unsigned int); 2122 int __generic_remap_file_range_prep(struct file *file_in, loff_t pos_in, 2123 struct file *file_out, loff_t pos_out, 2124 loff_t *len, unsigned int remap_flags, 2125 const struct iomap_ops *dax_read_ops); 2126 int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in, 2127 struct file *file_out, loff_t pos_out, 2128 loff_t *count, unsigned int remap_flags); 2129 extern loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in, 2130 struct file *file_out, loff_t pos_out, 2131 loff_t len, unsigned int remap_flags); 2132 extern int vfs_dedupe_file_range(struct file *file, 2133 struct file_dedupe_range *same); 2134 extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos, 2135 struct file *dst_file, loff_t dst_pos, 2136 loff_t len, unsigned int remap_flags); 2137 2138 /** 2139 * enum freeze_holder - holder of the freeze 2140 * @FREEZE_HOLDER_KERNEL: kernel wants to freeze or thaw filesystem 2141 * @FREEZE_HOLDER_USERSPACE: userspace wants to freeze or thaw filesystem 2142 * @FREEZE_MAY_NEST: whether nesting freeze and thaw requests is allowed 2143 * 2144 * Indicate who the owner of the freeze or thaw request is and whether 2145 * the freeze needs to be exclusive or can nest. 2146 * Without @FREEZE_MAY_NEST, multiple freeze and thaw requests from the 2147 * same holder aren't allowed. It is however allowed to hold a single 2148 * @FREEZE_HOLDER_USERSPACE and a single @FREEZE_HOLDER_KERNEL freeze at 2149 * the same time. This is relied upon by some filesystems during online 2150 * repair or similar. 2151 */ 2152 enum freeze_holder { 2153 FREEZE_HOLDER_KERNEL = (1U << 0), 2154 FREEZE_HOLDER_USERSPACE = (1U << 1), 2155 FREEZE_MAY_NEST = (1U << 2), 2156 }; 2157 2158 struct super_operations { 2159 struct inode *(*alloc_inode)(struct super_block *sb); 2160 void (*destroy_inode)(struct inode *); 2161 void (*free_inode)(struct inode *); 2162 2163 void (*dirty_inode) (struct inode *, int flags); 2164 int (*write_inode) (struct inode *, struct writeback_control *wbc); 2165 int (*drop_inode) (struct inode *); 2166 void (*evict_inode) (struct inode *); 2167 void (*put_super) (struct super_block *); 2168 int (*sync_fs)(struct super_block *sb, int wait); 2169 int (*freeze_super) (struct super_block *, enum freeze_holder who); 2170 int (*freeze_fs) (struct super_block *); 2171 int (*thaw_super) (struct super_block *, enum freeze_holder who); 2172 int (*unfreeze_fs) (struct super_block *); 2173 int (*statfs) (struct dentry *, struct kstatfs *); 2174 int (*remount_fs) (struct super_block *, int *, char *); 2175 void (*umount_begin) (struct super_block *); 2176 2177 int (*show_options)(struct seq_file *, struct dentry *); 2178 int (*show_devname)(struct seq_file *, struct dentry *); 2179 int (*show_path)(struct seq_file *, struct dentry *); 2180 int (*show_stats)(struct seq_file *, struct dentry *); 2181 #ifdef CONFIG_QUOTA 2182 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t); 2183 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t); 2184 struct dquot __rcu **(*get_dquots)(struct inode *); 2185 #endif 2186 long (*nr_cached_objects)(struct super_block *, 2187 struct shrink_control *); 2188 long (*free_cached_objects)(struct super_block *, 2189 struct shrink_control *); 2190 void (*shutdown)(struct super_block *sb); 2191 }; 2192 2193 /* 2194 * Inode flags - they have no relation to superblock flags now 2195 */ 2196 #define S_SYNC (1 << 0) /* Writes are synced at once */ 2197 #define S_NOATIME (1 << 1) /* Do not update access times */ 2198 #define S_APPEND (1 << 2) /* Append-only file */ 2199 #define S_IMMUTABLE (1 << 3) /* Immutable file */ 2200 #define S_DEAD (1 << 4) /* removed, but still open directory */ 2201 #define S_NOQUOTA (1 << 5) /* Inode is not counted to quota */ 2202 #define S_DIRSYNC (1 << 6) /* Directory modifications are synchronous */ 2203 #define S_NOCMTIME (1 << 7) /* Do not update file c/mtime */ 2204 #define S_SWAPFILE (1 << 8) /* Do not truncate: swapon got its bmaps */ 2205 #define S_PRIVATE (1 << 9) /* Inode is fs-internal */ 2206 #define S_IMA (1 << 10) /* Inode has an associated IMA struct */ 2207 #define S_AUTOMOUNT (1 << 11) /* Automount/referral quasi-directory */ 2208 #define S_NOSEC (1 << 12) /* no suid or xattr security attributes */ 2209 #ifdef CONFIG_FS_DAX 2210 #define S_DAX (1 << 13) /* Direct Access, avoiding the page cache */ 2211 #else 2212 #define S_DAX 0 /* Make all the DAX code disappear */ 2213 #endif 2214 #define S_ENCRYPTED (1 << 14) /* Encrypted file (using fs/crypto/) */ 2215 #define S_CASEFOLD (1 << 15) /* Casefolded file */ 2216 #define S_VERITY (1 << 16) /* Verity file (using fs/verity/) */ 2217 #define S_KERNEL_FILE (1 << 17) /* File is in use by the kernel (eg. fs/cachefiles) */ 2218 2219 /* 2220 * Note that nosuid etc flags are inode-specific: setting some file-system 2221 * flags just means all the inodes inherit those flags by default. It might be 2222 * possible to override it selectively if you really wanted to with some 2223 * ioctl() that is not currently implemented. 2224 * 2225 * Exception: SB_RDONLY is always applied to the entire file system. 2226 * 2227 * Unfortunately, it is possible to change a filesystems flags with it mounted 2228 * with files in use. This means that all of the inodes will not have their 2229 * i_flags updated. Hence, i_flags no longer inherit the superblock mount 2230 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org 2231 */ 2232 #define __IS_FLG(inode, flg) ((inode)->i_sb->s_flags & (flg)) 2233 2234 static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; } 2235 #define IS_RDONLY(inode) sb_rdonly((inode)->i_sb) 2236 #define IS_SYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS) || \ 2237 ((inode)->i_flags & S_SYNC)) 2238 #define IS_DIRSYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS|SB_DIRSYNC) || \ 2239 ((inode)->i_flags & (S_SYNC|S_DIRSYNC))) 2240 #define IS_MANDLOCK(inode) __IS_FLG(inode, SB_MANDLOCK) 2241 #define IS_NOATIME(inode) __IS_FLG(inode, SB_RDONLY|SB_NOATIME) 2242 #define IS_I_VERSION(inode) __IS_FLG(inode, SB_I_VERSION) 2243 2244 #define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA) 2245 #define IS_APPEND(inode) ((inode)->i_flags & S_APPEND) 2246 #define IS_IMMUTABLE(inode) ((inode)->i_flags & S_IMMUTABLE) 2247 2248 #ifdef CONFIG_FS_POSIX_ACL 2249 #define IS_POSIXACL(inode) __IS_FLG(inode, SB_POSIXACL) 2250 #else 2251 #define IS_POSIXACL(inode) 0 2252 #endif 2253 2254 #define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD) 2255 #define IS_NOCMTIME(inode) ((inode)->i_flags & S_NOCMTIME) 2256 #define IS_SWAPFILE(inode) ((inode)->i_flags & S_SWAPFILE) 2257 #define IS_PRIVATE(inode) ((inode)->i_flags & S_PRIVATE) 2258 #define IS_IMA(inode) ((inode)->i_flags & S_IMA) 2259 #define IS_AUTOMOUNT(inode) ((inode)->i_flags & S_AUTOMOUNT) 2260 #define IS_NOSEC(inode) ((inode)->i_flags & S_NOSEC) 2261 #define IS_DAX(inode) ((inode)->i_flags & S_DAX) 2262 #define IS_ENCRYPTED(inode) ((inode)->i_flags & S_ENCRYPTED) 2263 #define IS_CASEFOLDED(inode) ((inode)->i_flags & S_CASEFOLD) 2264 #define IS_VERITY(inode) ((inode)->i_flags & S_VERITY) 2265 2266 #define IS_WHITEOUT(inode) (S_ISCHR(inode->i_mode) && \ 2267 (inode)->i_rdev == WHITEOUT_DEV) 2268 2269 static inline bool HAS_UNMAPPED_ID(struct mnt_idmap *idmap, 2270 struct inode *inode) 2271 { 2272 return !vfsuid_valid(i_uid_into_vfsuid(idmap, inode)) || 2273 !vfsgid_valid(i_gid_into_vfsgid(idmap, inode)); 2274 } 2275 2276 static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp) 2277 { 2278 *kiocb = (struct kiocb) { 2279 .ki_filp = filp, 2280 .ki_flags = filp->f_iocb_flags, 2281 .ki_ioprio = get_current_ioprio(), 2282 }; 2283 } 2284 2285 static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src, 2286 struct file *filp) 2287 { 2288 *kiocb = (struct kiocb) { 2289 .ki_filp = filp, 2290 .ki_flags = kiocb_src->ki_flags, 2291 .ki_ioprio = kiocb_src->ki_ioprio, 2292 .ki_pos = kiocb_src->ki_pos, 2293 }; 2294 } 2295 2296 /* 2297 * Inode state bits. Protected by inode->i_lock 2298 * 2299 * Four bits determine the dirty state of the inode: I_DIRTY_SYNC, 2300 * I_DIRTY_DATASYNC, I_DIRTY_PAGES, and I_DIRTY_TIME. 2301 * 2302 * Four bits define the lifetime of an inode. Initially, inodes are I_NEW, 2303 * until that flag is cleared. I_WILL_FREE, I_FREEING and I_CLEAR are set at 2304 * various stages of removing an inode. 2305 * 2306 * Two bits are used for locking and completion notification, I_NEW and I_SYNC. 2307 * 2308 * I_DIRTY_SYNC Inode is dirty, but doesn't have to be written on 2309 * fdatasync() (unless I_DIRTY_DATASYNC is also set). 2310 * Timestamp updates are the usual cause. 2311 * I_DIRTY_DATASYNC Data-related inode changes pending. We keep track of 2312 * these changes separately from I_DIRTY_SYNC so that we 2313 * don't have to write inode on fdatasync() when only 2314 * e.g. the timestamps have changed. 2315 * I_DIRTY_PAGES Inode has dirty pages. Inode itself may be clean. 2316 * I_DIRTY_TIME The inode itself has dirty timestamps, and the 2317 * lazytime mount option is enabled. We keep track of this 2318 * separately from I_DIRTY_SYNC in order to implement 2319 * lazytime. This gets cleared if I_DIRTY_INODE 2320 * (I_DIRTY_SYNC and/or I_DIRTY_DATASYNC) gets set. But 2321 * I_DIRTY_TIME can still be set if I_DIRTY_SYNC is already 2322 * in place because writeback might already be in progress 2323 * and we don't want to lose the time update 2324 * I_NEW Serves as both a mutex and completion notification. 2325 * New inodes set I_NEW. If two processes both create 2326 * the same inode, one of them will release its inode and 2327 * wait for I_NEW to be released before returning. 2328 * Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can 2329 * also cause waiting on I_NEW, without I_NEW actually 2330 * being set. find_inode() uses this to prevent returning 2331 * nearly-dead inodes. 2332 * I_WILL_FREE Must be set when calling write_inode_now() if i_count 2333 * is zero. I_FREEING must be set when I_WILL_FREE is 2334 * cleared. 2335 * I_FREEING Set when inode is about to be freed but still has dirty 2336 * pages or buffers attached or the inode itself is still 2337 * dirty. 2338 * I_CLEAR Added by clear_inode(). In this state the inode is 2339 * clean and can be destroyed. Inode keeps I_FREEING. 2340 * 2341 * Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are 2342 * prohibited for many purposes. iget() must wait for 2343 * the inode to be completely released, then create it 2344 * anew. Other functions will just ignore such inodes, 2345 * if appropriate. I_NEW is used for waiting. 2346 * 2347 * I_SYNC Writeback of inode is running. The bit is set during 2348 * data writeback, and cleared with a wakeup on the bit 2349 * address once it is done. The bit is also used to pin 2350 * the inode in memory for flusher thread. 2351 * 2352 * I_REFERENCED Marks the inode as recently references on the LRU list. 2353 * 2354 * I_DIO_WAKEUP Never set. Only used as a key for wait_on_bit(). 2355 * 2356 * I_WB_SWITCH Cgroup bdi_writeback switching in progress. Used to 2357 * synchronize competing switching instances and to tell 2358 * wb stat updates to grab the i_pages lock. See 2359 * inode_switch_wbs_work_fn() for details. 2360 * 2361 * I_OVL_INUSE Used by overlayfs to get exclusive ownership on upper 2362 * and work dirs among overlayfs mounts. 2363 * 2364 * I_CREATING New object's inode in the middle of setting up. 2365 * 2366 * I_DONTCACHE Evict inode as soon as it is not used anymore. 2367 * 2368 * I_SYNC_QUEUED Inode is queued in b_io or b_more_io writeback lists. 2369 * Used to detect that mark_inode_dirty() should not move 2370 * inode between dirty lists. 2371 * 2372 * I_PINNING_FSCACHE_WB Inode is pinning an fscache object for writeback. 2373 * 2374 * Q: What is the difference between I_WILL_FREE and I_FREEING? 2375 */ 2376 #define I_DIRTY_SYNC (1 << 0) 2377 #define I_DIRTY_DATASYNC (1 << 1) 2378 #define I_DIRTY_PAGES (1 << 2) 2379 #define __I_NEW 3 2380 #define I_NEW (1 << __I_NEW) 2381 #define I_WILL_FREE (1 << 4) 2382 #define I_FREEING (1 << 5) 2383 #define I_CLEAR (1 << 6) 2384 #define __I_SYNC 7 2385 #define I_SYNC (1 << __I_SYNC) 2386 #define I_REFERENCED (1 << 8) 2387 #define __I_DIO_WAKEUP 9 2388 #define I_DIO_WAKEUP (1 << __I_DIO_WAKEUP) 2389 #define I_LINKABLE (1 << 10) 2390 #define I_DIRTY_TIME (1 << 11) 2391 #define I_WB_SWITCH (1 << 13) 2392 #define I_OVL_INUSE (1 << 14) 2393 #define I_CREATING (1 << 15) 2394 #define I_DONTCACHE (1 << 16) 2395 #define I_SYNC_QUEUED (1 << 17) 2396 #define I_PINNING_NETFS_WB (1 << 18) 2397 2398 #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC) 2399 #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES) 2400 #define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME) 2401 2402 extern void __mark_inode_dirty(struct inode *, int); 2403 static inline void mark_inode_dirty(struct inode *inode) 2404 { 2405 __mark_inode_dirty(inode, I_DIRTY); 2406 } 2407 2408 static inline void mark_inode_dirty_sync(struct inode *inode) 2409 { 2410 __mark_inode_dirty(inode, I_DIRTY_SYNC); 2411 } 2412 2413 /* 2414 * Returns true if the given inode itself only has dirty timestamps (its pages 2415 * may still be dirty) and isn't currently being allocated or freed. 2416 * Filesystems should call this if when writing an inode when lazytime is 2417 * enabled, they want to opportunistically write the timestamps of other inodes 2418 * located very nearby on-disk, e.g. in the same inode block. This returns true 2419 * if the given inode is in need of such an opportunistic update. Requires 2420 * i_lock, or at least later re-checking under i_lock. 2421 */ 2422 static inline bool inode_is_dirtytime_only(struct inode *inode) 2423 { 2424 return (inode->i_state & (I_DIRTY_TIME | I_NEW | 2425 I_FREEING | I_WILL_FREE)) == I_DIRTY_TIME; 2426 } 2427 2428 extern void inc_nlink(struct inode *inode); 2429 extern void drop_nlink(struct inode *inode); 2430 extern void clear_nlink(struct inode *inode); 2431 extern void set_nlink(struct inode *inode, unsigned int nlink); 2432 2433 static inline void inode_inc_link_count(struct inode *inode) 2434 { 2435 inc_nlink(inode); 2436 mark_inode_dirty(inode); 2437 } 2438 2439 static inline void inode_dec_link_count(struct inode *inode) 2440 { 2441 drop_nlink(inode); 2442 mark_inode_dirty(inode); 2443 } 2444 2445 enum file_time_flags { 2446 S_ATIME = 1, 2447 S_MTIME = 2, 2448 S_CTIME = 4, 2449 S_VERSION = 8, 2450 }; 2451 2452 extern bool atime_needs_update(const struct path *, struct inode *); 2453 extern void touch_atime(const struct path *); 2454 int inode_update_time(struct inode *inode, int flags); 2455 2456 static inline void file_accessed(struct file *file) 2457 { 2458 if (!(file->f_flags & O_NOATIME)) 2459 touch_atime(&file->f_path); 2460 } 2461 2462 extern int file_modified(struct file *file); 2463 int kiocb_modified(struct kiocb *iocb); 2464 2465 int sync_inode_metadata(struct inode *inode, int wait); 2466 2467 struct file_system_type { 2468 const char *name; 2469 int fs_flags; 2470 #define FS_REQUIRES_DEV 1 2471 #define FS_BINARY_MOUNTDATA 2 2472 #define FS_HAS_SUBTYPE 4 2473 #define FS_USERNS_MOUNT 8 /* Can be mounted by userns root */ 2474 #define FS_DISALLOW_NOTIFY_PERM 16 /* Disable fanotify permission events */ 2475 #define FS_ALLOW_IDMAP 32 /* FS has been updated to handle vfs idmappings. */ 2476 #define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() during rename() internally. */ 2477 int (*init_fs_context)(struct fs_context *); 2478 const struct fs_parameter_spec *parameters; 2479 struct dentry *(*mount) (struct file_system_type *, int, 2480 const char *, void *); 2481 void (*kill_sb) (struct super_block *); 2482 struct module *owner; 2483 struct file_system_type * next; 2484 struct hlist_head fs_supers; 2485 2486 struct lock_class_key s_lock_key; 2487 struct lock_class_key s_umount_key; 2488 struct lock_class_key s_vfs_rename_key; 2489 struct lock_class_key s_writers_key[SB_FREEZE_LEVELS]; 2490 2491 struct lock_class_key i_lock_key; 2492 struct lock_class_key i_mutex_key; 2493 struct lock_class_key invalidate_lock_key; 2494 struct lock_class_key i_mutex_dir_key; 2495 }; 2496 2497 #define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME) 2498 2499 extern struct dentry *mount_bdev(struct file_system_type *fs_type, 2500 int flags, const char *dev_name, void *data, 2501 int (*fill_super)(struct super_block *, void *, int)); 2502 extern struct dentry *mount_single(struct file_system_type *fs_type, 2503 int flags, void *data, 2504 int (*fill_super)(struct super_block *, void *, int)); 2505 extern struct dentry *mount_nodev(struct file_system_type *fs_type, 2506 int flags, void *data, 2507 int (*fill_super)(struct super_block *, void *, int)); 2508 extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path); 2509 void retire_super(struct super_block *sb); 2510 void generic_shutdown_super(struct super_block *sb); 2511 void kill_block_super(struct super_block *sb); 2512 void kill_anon_super(struct super_block *sb); 2513 void kill_litter_super(struct super_block *sb); 2514 void deactivate_super(struct super_block *sb); 2515 void deactivate_locked_super(struct super_block *sb); 2516 int set_anon_super(struct super_block *s, void *data); 2517 int set_anon_super_fc(struct super_block *s, struct fs_context *fc); 2518 int get_anon_bdev(dev_t *); 2519 void free_anon_bdev(dev_t); 2520 struct super_block *sget_fc(struct fs_context *fc, 2521 int (*test)(struct super_block *, struct fs_context *), 2522 int (*set)(struct super_block *, struct fs_context *)); 2523 struct super_block *sget(struct file_system_type *type, 2524 int (*test)(struct super_block *,void *), 2525 int (*set)(struct super_block *,void *), 2526 int flags, void *data); 2527 struct super_block *sget_dev(struct fs_context *fc, dev_t dev); 2528 2529 /* Alas, no aliases. Too much hassle with bringing module.h everywhere */ 2530 #define fops_get(fops) \ 2531 (((fops) && try_module_get((fops)->owner) ? (fops) : NULL)) 2532 #define fops_put(fops) \ 2533 do { if (fops) module_put((fops)->owner); } while(0) 2534 /* 2535 * This one is to be used *ONLY* from ->open() instances. 2536 * fops must be non-NULL, pinned down *and* module dependencies 2537 * should be sufficient to pin the caller down as well. 2538 */ 2539 #define replace_fops(f, fops) \ 2540 do { \ 2541 struct file *__file = (f); \ 2542 fops_put(__file->f_op); \ 2543 BUG_ON(!(__file->f_op = (fops))); \ 2544 } while(0) 2545 2546 extern int register_filesystem(struct file_system_type *); 2547 extern int unregister_filesystem(struct file_system_type *); 2548 extern int vfs_statfs(const struct path *, struct kstatfs *); 2549 extern int user_statfs(const char __user *, struct kstatfs *); 2550 extern int fd_statfs(int, struct kstatfs *); 2551 int freeze_super(struct super_block *super, enum freeze_holder who); 2552 int thaw_super(struct super_block *super, enum freeze_holder who); 2553 extern __printf(2, 3) 2554 int super_setup_bdi_name(struct super_block *sb, char *fmt, ...); 2555 extern int super_setup_bdi(struct super_block *sb); 2556 2557 static inline void super_set_uuid(struct super_block *sb, const u8 *uuid, unsigned len) 2558 { 2559 if (WARN_ON(len > sizeof(sb->s_uuid))) 2560 len = sizeof(sb->s_uuid); 2561 sb->s_uuid_len = len; 2562 memcpy(&sb->s_uuid, uuid, len); 2563 } 2564 2565 /* set sb sysfs name based on sb->s_bdev */ 2566 static inline void super_set_sysfs_name_bdev(struct super_block *sb) 2567 { 2568 snprintf(sb->s_sysfs_name, sizeof(sb->s_sysfs_name), "%pg", sb->s_bdev); 2569 } 2570 2571 /* set sb sysfs name based on sb->s_uuid */ 2572 static inline void super_set_sysfs_name_uuid(struct super_block *sb) 2573 { 2574 WARN_ON(sb->s_uuid_len != sizeof(sb->s_uuid)); 2575 snprintf(sb->s_sysfs_name, sizeof(sb->s_sysfs_name), "%pU", sb->s_uuid.b); 2576 } 2577 2578 /* set sb sysfs name based on sb->s_id */ 2579 static inline void super_set_sysfs_name_id(struct super_block *sb) 2580 { 2581 strscpy(sb->s_sysfs_name, sb->s_id, sizeof(sb->s_sysfs_name)); 2582 } 2583 2584 /* try to use something standard before you use this */ 2585 __printf(2, 3) 2586 static inline void super_set_sysfs_name_generic(struct super_block *sb, const char *fmt, ...) 2587 { 2588 va_list args; 2589 2590 va_start(args, fmt); 2591 vsnprintf(sb->s_sysfs_name, sizeof(sb->s_sysfs_name), fmt, args); 2592 va_end(args); 2593 } 2594 2595 extern int current_umask(void); 2596 2597 extern void ihold(struct inode * inode); 2598 extern void iput(struct inode *); 2599 int inode_update_timestamps(struct inode *inode, int flags); 2600 int generic_update_time(struct inode *, int); 2601 2602 /* /sys/fs */ 2603 extern struct kobject *fs_kobj; 2604 2605 #define MAX_RW_COUNT (INT_MAX & PAGE_MASK) 2606 2607 /* fs/open.c */ 2608 struct audit_names; 2609 struct filename { 2610 const char *name; /* pointer to actual string */ 2611 const __user char *uptr; /* original userland pointer */ 2612 atomic_t refcnt; 2613 struct audit_names *aname; 2614 const char iname[]; 2615 }; 2616 static_assert(offsetof(struct filename, iname) % sizeof(long) == 0); 2617 2618 static inline struct mnt_idmap *file_mnt_idmap(const struct file *file) 2619 { 2620 return mnt_idmap(file->f_path.mnt); 2621 } 2622 2623 /** 2624 * is_idmapped_mnt - check whether a mount is mapped 2625 * @mnt: the mount to check 2626 * 2627 * If @mnt has an non @nop_mnt_idmap attached to it then @mnt is mapped. 2628 * 2629 * Return: true if mount is mapped, false if not. 2630 */ 2631 static inline bool is_idmapped_mnt(const struct vfsmount *mnt) 2632 { 2633 return mnt_idmap(mnt) != &nop_mnt_idmap; 2634 } 2635 2636 extern long vfs_truncate(const struct path *, loff_t); 2637 int do_truncate(struct mnt_idmap *, struct dentry *, loff_t start, 2638 unsigned int time_attrs, struct file *filp); 2639 extern int vfs_fallocate(struct file *file, int mode, loff_t offset, 2640 loff_t len); 2641 extern long do_sys_open(int dfd, const char __user *filename, int flags, 2642 umode_t mode); 2643 extern struct file *file_open_name(struct filename *, int, umode_t); 2644 extern struct file *filp_open(const char *, int, umode_t); 2645 extern struct file *file_open_root(const struct path *, 2646 const char *, int, umode_t); 2647 static inline struct file *file_open_root_mnt(struct vfsmount *mnt, 2648 const char *name, int flags, umode_t mode) 2649 { 2650 return file_open_root(&(struct path){.mnt = mnt, .dentry = mnt->mnt_root}, 2651 name, flags, mode); 2652 } 2653 struct file *dentry_open(const struct path *path, int flags, 2654 const struct cred *creds); 2655 struct file *dentry_create(const struct path *path, int flags, umode_t mode, 2656 const struct cred *cred); 2657 struct path *backing_file_user_path(struct file *f); 2658 2659 /* 2660 * When mmapping a file on a stackable filesystem (e.g., overlayfs), the file 2661 * stored in ->vm_file is a backing file whose f_inode is on the underlying 2662 * filesystem. When the mapped file path and inode number are displayed to 2663 * user (e.g. via /proc/<pid>/maps), these helpers should be used to get the 2664 * path and inode number to display to the user, which is the path of the fd 2665 * that user has requested to map and the inode number that would be returned 2666 * by fstat() on that same fd. 2667 */ 2668 /* Get the path to display in /proc/<pid>/maps */ 2669 static inline const struct path *file_user_path(struct file *f) 2670 { 2671 if (unlikely(f->f_mode & FMODE_BACKING)) 2672 return backing_file_user_path(f); 2673 return &f->f_path; 2674 } 2675 /* Get the inode whose inode number to display in /proc/<pid>/maps */ 2676 static inline const struct inode *file_user_inode(struct file *f) 2677 { 2678 if (unlikely(f->f_mode & FMODE_BACKING)) 2679 return d_inode(backing_file_user_path(f)->dentry); 2680 return file_inode(f); 2681 } 2682 2683 static inline struct file *file_clone_open(struct file *file) 2684 { 2685 return dentry_open(&file->f_path, file->f_flags, file->f_cred); 2686 } 2687 extern int filp_close(struct file *, fl_owner_t id); 2688 2689 extern struct filename *getname_flags(const char __user *, int, int *); 2690 extern struct filename *getname_uflags(const char __user *, int); 2691 extern struct filename *getname(const char __user *); 2692 extern struct filename *getname_kernel(const char *); 2693 extern void putname(struct filename *name); 2694 2695 extern int finish_open(struct file *file, struct dentry *dentry, 2696 int (*open)(struct inode *, struct file *)); 2697 extern int finish_no_open(struct file *file, struct dentry *dentry); 2698 2699 /* Helper for the simple case when original dentry is used */ 2700 static inline int finish_open_simple(struct file *file, int error) 2701 { 2702 if (error) 2703 return error; 2704 2705 return finish_open(file, file->f_path.dentry, NULL); 2706 } 2707 2708 /* fs/dcache.c */ 2709 extern void __init vfs_caches_init_early(void); 2710 extern void __init vfs_caches_init(void); 2711 2712 extern struct kmem_cache *names_cachep; 2713 2714 #define __getname() kmem_cache_alloc(names_cachep, GFP_KERNEL) 2715 #define __putname(name) kmem_cache_free(names_cachep, (void *)(name)) 2716 2717 extern struct super_block *blockdev_superblock; 2718 static inline bool sb_is_blkdev_sb(struct super_block *sb) 2719 { 2720 return IS_ENABLED(CONFIG_BLOCK) && sb == blockdev_superblock; 2721 } 2722 2723 void emergency_thaw_all(void); 2724 extern int sync_filesystem(struct super_block *); 2725 extern const struct file_operations def_blk_fops; 2726 extern const struct file_operations def_chr_fops; 2727 2728 /* fs/char_dev.c */ 2729 #define CHRDEV_MAJOR_MAX 512 2730 /* Marks the bottom of the first segment of free char majors */ 2731 #define CHRDEV_MAJOR_DYN_END 234 2732 /* Marks the top and bottom of the second segment of free char majors */ 2733 #define CHRDEV_MAJOR_DYN_EXT_START 511 2734 #define CHRDEV_MAJOR_DYN_EXT_END 384 2735 2736 extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *); 2737 extern int register_chrdev_region(dev_t, unsigned, const char *); 2738 extern int __register_chrdev(unsigned int major, unsigned int baseminor, 2739 unsigned int count, const char *name, 2740 const struct file_operations *fops); 2741 extern void __unregister_chrdev(unsigned int major, unsigned int baseminor, 2742 unsigned int count, const char *name); 2743 extern void unregister_chrdev_region(dev_t, unsigned); 2744 extern void chrdev_show(struct seq_file *,off_t); 2745 2746 static inline int register_chrdev(unsigned int major, const char *name, 2747 const struct file_operations *fops) 2748 { 2749 return __register_chrdev(major, 0, 256, name, fops); 2750 } 2751 2752 static inline void unregister_chrdev(unsigned int major, const char *name) 2753 { 2754 __unregister_chrdev(major, 0, 256, name); 2755 } 2756 2757 extern void init_special_inode(struct inode *, umode_t, dev_t); 2758 2759 /* Invalid inode operations -- fs/bad_inode.c */ 2760 extern void make_bad_inode(struct inode *); 2761 extern bool is_bad_inode(struct inode *); 2762 2763 extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart, 2764 loff_t lend); 2765 extern int __must_check file_check_and_advance_wb_err(struct file *file); 2766 extern int __must_check file_write_and_wait_range(struct file *file, 2767 loff_t start, loff_t end); 2768 2769 static inline int file_write_and_wait(struct file *file) 2770 { 2771 return file_write_and_wait_range(file, 0, LLONG_MAX); 2772 } 2773 2774 extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end, 2775 int datasync); 2776 extern int vfs_fsync(struct file *file, int datasync); 2777 2778 extern int sync_file_range(struct file *file, loff_t offset, loff_t nbytes, 2779 unsigned int flags); 2780 2781 static inline bool iocb_is_dsync(const struct kiocb *iocb) 2782 { 2783 return (iocb->ki_flags & IOCB_DSYNC) || 2784 IS_SYNC(iocb->ki_filp->f_mapping->host); 2785 } 2786 2787 /* 2788 * Sync the bytes written if this was a synchronous write. Expect ki_pos 2789 * to already be updated for the write, and will return either the amount 2790 * of bytes passed in, or an error if syncing the file failed. 2791 */ 2792 static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count) 2793 { 2794 if (iocb_is_dsync(iocb)) { 2795 int ret = vfs_fsync_range(iocb->ki_filp, 2796 iocb->ki_pos - count, iocb->ki_pos - 1, 2797 (iocb->ki_flags & IOCB_SYNC) ? 0 : 1); 2798 if (ret) 2799 return ret; 2800 } 2801 2802 return count; 2803 } 2804 2805 extern void emergency_sync(void); 2806 extern void emergency_remount(void); 2807 2808 #ifdef CONFIG_BLOCK 2809 extern int bmap(struct inode *inode, sector_t *block); 2810 #else 2811 static inline int bmap(struct inode *inode, sector_t *block) 2812 { 2813 return -EINVAL; 2814 } 2815 #endif 2816 2817 int notify_change(struct mnt_idmap *, struct dentry *, 2818 struct iattr *, struct inode **); 2819 int inode_permission(struct mnt_idmap *, struct inode *, int); 2820 int generic_permission(struct mnt_idmap *, struct inode *, int); 2821 static inline int file_permission(struct file *file, int mask) 2822 { 2823 return inode_permission(file_mnt_idmap(file), 2824 file_inode(file), mask); 2825 } 2826 static inline int path_permission(const struct path *path, int mask) 2827 { 2828 return inode_permission(mnt_idmap(path->mnt), 2829 d_inode(path->dentry), mask); 2830 } 2831 int __check_sticky(struct mnt_idmap *idmap, struct inode *dir, 2832 struct inode *inode); 2833 2834 static inline bool execute_ok(struct inode *inode) 2835 { 2836 return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode); 2837 } 2838 2839 static inline bool inode_wrong_type(const struct inode *inode, umode_t mode) 2840 { 2841 return (inode->i_mode ^ mode) & S_IFMT; 2842 } 2843 2844 /** 2845 * file_start_write - get write access to a superblock for regular file io 2846 * @file: the file we want to write to 2847 * 2848 * This is a variant of sb_start_write() which is a noop on non-regualr file. 2849 * Should be matched with a call to file_end_write(). 2850 */ 2851 static inline void file_start_write(struct file *file) 2852 { 2853 if (!S_ISREG(file_inode(file)->i_mode)) 2854 return; 2855 sb_start_write(file_inode(file)->i_sb); 2856 } 2857 2858 static inline bool file_start_write_trylock(struct file *file) 2859 { 2860 if (!S_ISREG(file_inode(file)->i_mode)) 2861 return true; 2862 return sb_start_write_trylock(file_inode(file)->i_sb); 2863 } 2864 2865 /** 2866 * file_end_write - drop write access to a superblock of a regular file 2867 * @file: the file we wrote to 2868 * 2869 * Should be matched with a call to file_start_write(). 2870 */ 2871 static inline void file_end_write(struct file *file) 2872 { 2873 if (!S_ISREG(file_inode(file)->i_mode)) 2874 return; 2875 sb_end_write(file_inode(file)->i_sb); 2876 } 2877 2878 /** 2879 * kiocb_start_write - get write access to a superblock for async file io 2880 * @iocb: the io context we want to submit the write with 2881 * 2882 * This is a variant of sb_start_write() for async io submission. 2883 * Should be matched with a call to kiocb_end_write(). 2884 */ 2885 static inline void kiocb_start_write(struct kiocb *iocb) 2886 { 2887 struct inode *inode = file_inode(iocb->ki_filp); 2888 2889 sb_start_write(inode->i_sb); 2890 /* 2891 * Fool lockdep by telling it the lock got released so that it 2892 * doesn't complain about the held lock when we return to userspace. 2893 */ 2894 __sb_writers_release(inode->i_sb, SB_FREEZE_WRITE); 2895 } 2896 2897 /** 2898 * kiocb_end_write - drop write access to a superblock after async file io 2899 * @iocb: the io context we sumbitted the write with 2900 * 2901 * Should be matched with a call to kiocb_start_write(). 2902 */ 2903 static inline void kiocb_end_write(struct kiocb *iocb) 2904 { 2905 struct inode *inode = file_inode(iocb->ki_filp); 2906 2907 /* 2908 * Tell lockdep we inherited freeze protection from submission thread. 2909 */ 2910 __sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE); 2911 sb_end_write(inode->i_sb); 2912 } 2913 2914 /* 2915 * This is used for regular files where some users -- especially the 2916 * currently executed binary in a process, previously handled via 2917 * VM_DENYWRITE -- cannot handle concurrent write (and maybe mmap 2918 * read-write shared) accesses. 2919 * 2920 * get_write_access() gets write permission for a file. 2921 * put_write_access() releases this write permission. 2922 * deny_write_access() denies write access to a file. 2923 * allow_write_access() re-enables write access to a file. 2924 * 2925 * The i_writecount field of an inode can have the following values: 2926 * 0: no write access, no denied write access 2927 * < 0: (-i_writecount) users that denied write access to the file. 2928 * > 0: (i_writecount) users that have write access to the file. 2929 * 2930 * Normally we operate on that counter with atomic_{inc,dec} and it's safe 2931 * except for the cases where we don't hold i_writecount yet. Then we need to 2932 * use {get,deny}_write_access() - these functions check the sign and refuse 2933 * to do the change if sign is wrong. 2934 */ 2935 static inline int get_write_access(struct inode *inode) 2936 { 2937 return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY; 2938 } 2939 static inline int deny_write_access(struct file *file) 2940 { 2941 struct inode *inode = file_inode(file); 2942 return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY; 2943 } 2944 static inline void put_write_access(struct inode * inode) 2945 { 2946 atomic_dec(&inode->i_writecount); 2947 } 2948 static inline void allow_write_access(struct file *file) 2949 { 2950 if (file) 2951 atomic_inc(&file_inode(file)->i_writecount); 2952 } 2953 static inline bool inode_is_open_for_write(const struct inode *inode) 2954 { 2955 return atomic_read(&inode->i_writecount) > 0; 2956 } 2957 2958 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING) 2959 static inline void i_readcount_dec(struct inode *inode) 2960 { 2961 BUG_ON(atomic_dec_return(&inode->i_readcount) < 0); 2962 } 2963 static inline void i_readcount_inc(struct inode *inode) 2964 { 2965 atomic_inc(&inode->i_readcount); 2966 } 2967 #else 2968 static inline void i_readcount_dec(struct inode *inode) 2969 { 2970 return; 2971 } 2972 static inline void i_readcount_inc(struct inode *inode) 2973 { 2974 return; 2975 } 2976 #endif 2977 extern int do_pipe_flags(int *, int); 2978 2979 extern ssize_t kernel_read(struct file *, void *, size_t, loff_t *); 2980 ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos); 2981 extern ssize_t kernel_write(struct file *, const void *, size_t, loff_t *); 2982 extern ssize_t __kernel_write(struct file *, const void *, size_t, loff_t *); 2983 extern struct file * open_exec(const char *); 2984 2985 /* fs/dcache.c -- generic fs support functions */ 2986 extern bool is_subdir(struct dentry *, struct dentry *); 2987 extern bool path_is_under(const struct path *, const struct path *); 2988 2989 extern char *file_path(struct file *, char *, int); 2990 2991 /** 2992 * is_dot_dotdot - returns true only if @name is "." or ".." 2993 * @name: file name to check 2994 * @len: length of file name, in bytes 2995 */ 2996 static inline bool is_dot_dotdot(const char *name, size_t len) 2997 { 2998 return len && unlikely(name[0] == '.') && 2999 (len == 1 || (len == 2 && name[1] == '.')); 3000 } 3001 3002 #include <linux/err.h> 3003 3004 /* needed for stackable file system support */ 3005 extern loff_t default_llseek(struct file *file, loff_t offset, int whence); 3006 3007 extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence); 3008 3009 extern int inode_init_always(struct super_block *, struct inode *); 3010 extern void inode_init_once(struct inode *); 3011 extern void address_space_init_once(struct address_space *mapping); 3012 extern struct inode * igrab(struct inode *); 3013 extern ino_t iunique(struct super_block *, ino_t); 3014 extern int inode_needs_sync(struct inode *inode); 3015 extern int generic_delete_inode(struct inode *inode); 3016 static inline int generic_drop_inode(struct inode *inode) 3017 { 3018 return !inode->i_nlink || inode_unhashed(inode); 3019 } 3020 extern void d_mark_dontcache(struct inode *inode); 3021 3022 extern struct inode *ilookup5_nowait(struct super_block *sb, 3023 unsigned long hashval, int (*test)(struct inode *, void *), 3024 void *data); 3025 extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval, 3026 int (*test)(struct inode *, void *), void *data); 3027 extern struct inode *ilookup(struct super_block *sb, unsigned long ino); 3028 3029 extern struct inode *inode_insert5(struct inode *inode, unsigned long hashval, 3030 int (*test)(struct inode *, void *), 3031 int (*set)(struct inode *, void *), 3032 void *data); 3033 extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *); 3034 extern struct inode * iget_locked(struct super_block *, unsigned long); 3035 extern struct inode *find_inode_nowait(struct super_block *, 3036 unsigned long, 3037 int (*match)(struct inode *, 3038 unsigned long, void *), 3039 void *data); 3040 extern struct inode *find_inode_rcu(struct super_block *, unsigned long, 3041 int (*)(struct inode *, void *), void *); 3042 extern struct inode *find_inode_by_ino_rcu(struct super_block *, unsigned long); 3043 extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *); 3044 extern int insert_inode_locked(struct inode *); 3045 #ifdef CONFIG_DEBUG_LOCK_ALLOC 3046 extern void lockdep_annotate_inode_mutex_key(struct inode *inode); 3047 #else 3048 static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { }; 3049 #endif 3050 extern void unlock_new_inode(struct inode *); 3051 extern void discard_new_inode(struct inode *); 3052 extern unsigned int get_next_ino(void); 3053 extern void evict_inodes(struct super_block *sb); 3054 void dump_mapping(const struct address_space *); 3055 3056 /* 3057 * Userspace may rely on the inode number being non-zero. For example, glibc 3058 * simply ignores files with zero i_ino in unlink() and other places. 3059 * 3060 * As an additional complication, if userspace was compiled with 3061 * _FILE_OFFSET_BITS=32 on a 64-bit kernel we'll only end up reading out the 3062 * lower 32 bits, so we need to check that those aren't zero explicitly. With 3063 * _FILE_OFFSET_BITS=64, this may cause some harmless false-negatives, but 3064 * better safe than sorry. 3065 */ 3066 static inline bool is_zero_ino(ino_t ino) 3067 { 3068 return (u32)ino == 0; 3069 } 3070 3071 extern void __iget(struct inode * inode); 3072 extern void iget_failed(struct inode *); 3073 extern void clear_inode(struct inode *); 3074 extern void __destroy_inode(struct inode *); 3075 extern struct inode *new_inode_pseudo(struct super_block *sb); 3076 extern struct inode *new_inode(struct super_block *sb); 3077 extern void free_inode_nonrcu(struct inode *inode); 3078 extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *); 3079 extern int file_remove_privs_flags(struct file *file, unsigned int flags); 3080 extern int file_remove_privs(struct file *); 3081 int setattr_should_drop_sgid(struct mnt_idmap *idmap, 3082 const struct inode *inode); 3083 3084 /* 3085 * This must be used for allocating filesystems specific inodes to set 3086 * up the inode reclaim context correctly. 3087 */ 3088 static inline void * 3089 alloc_inode_sb(struct super_block *sb, struct kmem_cache *cache, gfp_t gfp) 3090 { 3091 return kmem_cache_alloc_lru(cache, &sb->s_inode_lru, gfp); 3092 } 3093 3094 extern void __insert_inode_hash(struct inode *, unsigned long hashval); 3095 static inline void insert_inode_hash(struct inode *inode) 3096 { 3097 __insert_inode_hash(inode, inode->i_ino); 3098 } 3099 3100 extern void __remove_inode_hash(struct inode *); 3101 static inline void remove_inode_hash(struct inode *inode) 3102 { 3103 if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash)) 3104 __remove_inode_hash(inode); 3105 } 3106 3107 extern void inode_sb_list_add(struct inode *inode); 3108 extern void inode_add_lru(struct inode *inode); 3109 3110 extern int sb_set_blocksize(struct super_block *, int); 3111 extern int sb_min_blocksize(struct super_block *, int); 3112 3113 extern int generic_file_mmap(struct file *, struct vm_area_struct *); 3114 extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *); 3115 extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *); 3116 int generic_write_checks_count(struct kiocb *iocb, loff_t *count); 3117 extern int generic_write_check_limits(struct file *file, loff_t pos, 3118 loff_t *count); 3119 extern int generic_file_rw_checks(struct file *file_in, struct file *file_out); 3120 ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *to, 3121 ssize_t already_read); 3122 extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *); 3123 extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *); 3124 extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *); 3125 extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *); 3126 ssize_t generic_perform_write(struct kiocb *, struct iov_iter *); 3127 ssize_t direct_write_fallback(struct kiocb *iocb, struct iov_iter *iter, 3128 ssize_t direct_written, ssize_t buffered_written); 3129 3130 ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos, 3131 rwf_t flags); 3132 ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos, 3133 rwf_t flags); 3134 ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb, 3135 struct iov_iter *iter); 3136 ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb, 3137 struct iov_iter *iter); 3138 3139 /* fs/splice.c */ 3140 ssize_t filemap_splice_read(struct file *in, loff_t *ppos, 3141 struct pipe_inode_info *pipe, 3142 size_t len, unsigned int flags); 3143 ssize_t copy_splice_read(struct file *in, loff_t *ppos, 3144 struct pipe_inode_info *pipe, 3145 size_t len, unsigned int flags); 3146 extern ssize_t iter_file_splice_write(struct pipe_inode_info *, 3147 struct file *, loff_t *, size_t, unsigned int); 3148 3149 3150 extern void 3151 file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping); 3152 extern loff_t noop_llseek(struct file *file, loff_t offset, int whence); 3153 #define no_llseek NULL 3154 extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize); 3155 extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence); 3156 extern loff_t generic_file_llseek_size(struct file *file, loff_t offset, 3157 int whence, loff_t maxsize, loff_t eof); 3158 extern loff_t fixed_size_llseek(struct file *file, loff_t offset, 3159 int whence, loff_t size); 3160 extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t); 3161 extern loff_t no_seek_end_llseek(struct file *, loff_t, int); 3162 int rw_verify_area(int, struct file *, const loff_t *, size_t); 3163 extern int generic_file_open(struct inode * inode, struct file * filp); 3164 extern int nonseekable_open(struct inode * inode, struct file * filp); 3165 extern int stream_open(struct inode * inode, struct file * filp); 3166 3167 #ifdef CONFIG_BLOCK 3168 typedef void (dio_submit_t)(struct bio *bio, struct inode *inode, 3169 loff_t file_offset); 3170 3171 enum { 3172 /* need locking between buffered and direct access */ 3173 DIO_LOCKING = 0x01, 3174 3175 /* filesystem does not support filling holes */ 3176 DIO_SKIP_HOLES = 0x02, 3177 }; 3178 3179 ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode, 3180 struct block_device *bdev, struct iov_iter *iter, 3181 get_block_t get_block, 3182 dio_iodone_t end_io, 3183 int flags); 3184 3185 static inline ssize_t blockdev_direct_IO(struct kiocb *iocb, 3186 struct inode *inode, 3187 struct iov_iter *iter, 3188 get_block_t get_block) 3189 { 3190 return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter, 3191 get_block, NULL, DIO_LOCKING | DIO_SKIP_HOLES); 3192 } 3193 #endif 3194 3195 void inode_dio_wait(struct inode *inode); 3196 3197 /** 3198 * inode_dio_begin - signal start of a direct I/O requests 3199 * @inode: inode the direct I/O happens on 3200 * 3201 * This is called once we've finished processing a direct I/O request, 3202 * and is used to wake up callers waiting for direct I/O to be quiesced. 3203 */ 3204 static inline void inode_dio_begin(struct inode *inode) 3205 { 3206 atomic_inc(&inode->i_dio_count); 3207 } 3208 3209 /** 3210 * inode_dio_end - signal finish of a direct I/O requests 3211 * @inode: inode the direct I/O happens on 3212 * 3213 * This is called once we've finished processing a direct I/O request, 3214 * and is used to wake up callers waiting for direct I/O to be quiesced. 3215 */ 3216 static inline void inode_dio_end(struct inode *inode) 3217 { 3218 if (atomic_dec_and_test(&inode->i_dio_count)) 3219 wake_up_bit(&inode->i_state, __I_DIO_WAKEUP); 3220 } 3221 3222 extern void inode_set_flags(struct inode *inode, unsigned int flags, 3223 unsigned int mask); 3224 3225 extern const struct file_operations generic_ro_fops; 3226 3227 #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m)) 3228 3229 extern int readlink_copy(char __user *, int, const char *); 3230 extern int page_readlink(struct dentry *, char __user *, int); 3231 extern const char *page_get_link(struct dentry *, struct inode *, 3232 struct delayed_call *); 3233 extern void page_put_link(void *); 3234 extern int page_symlink(struct inode *inode, const char *symname, int len); 3235 extern const struct inode_operations page_symlink_inode_operations; 3236 extern void kfree_link(void *); 3237 void generic_fillattr(struct mnt_idmap *, u32, struct inode *, struct kstat *); 3238 void generic_fill_statx_attr(struct inode *inode, struct kstat *stat); 3239 extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int); 3240 extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int); 3241 void __inode_add_bytes(struct inode *inode, loff_t bytes); 3242 void inode_add_bytes(struct inode *inode, loff_t bytes); 3243 void __inode_sub_bytes(struct inode *inode, loff_t bytes); 3244 void inode_sub_bytes(struct inode *inode, loff_t bytes); 3245 static inline loff_t __inode_get_bytes(struct inode *inode) 3246 { 3247 return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes; 3248 } 3249 loff_t inode_get_bytes(struct inode *inode); 3250 void inode_set_bytes(struct inode *inode, loff_t bytes); 3251 const char *simple_get_link(struct dentry *, struct inode *, 3252 struct delayed_call *); 3253 extern const struct inode_operations simple_symlink_inode_operations; 3254 3255 extern int iterate_dir(struct file *, struct dir_context *); 3256 3257 int vfs_fstatat(int dfd, const char __user *filename, struct kstat *stat, 3258 int flags); 3259 int vfs_fstat(int fd, struct kstat *stat); 3260 3261 static inline int vfs_stat(const char __user *filename, struct kstat *stat) 3262 { 3263 return vfs_fstatat(AT_FDCWD, filename, stat, 0); 3264 } 3265 static inline int vfs_lstat(const char __user *name, struct kstat *stat) 3266 { 3267 return vfs_fstatat(AT_FDCWD, name, stat, AT_SYMLINK_NOFOLLOW); 3268 } 3269 3270 extern const char *vfs_get_link(struct dentry *, struct delayed_call *); 3271 extern int vfs_readlink(struct dentry *, char __user *, int); 3272 3273 extern struct file_system_type *get_filesystem(struct file_system_type *fs); 3274 extern void put_filesystem(struct file_system_type *fs); 3275 extern struct file_system_type *get_fs_type(const char *name); 3276 extern void drop_super(struct super_block *sb); 3277 extern void drop_super_exclusive(struct super_block *sb); 3278 extern void iterate_supers(void (*)(struct super_block *, void *), void *); 3279 extern void iterate_supers_type(struct file_system_type *, 3280 void (*)(struct super_block *, void *), void *); 3281 3282 extern int dcache_dir_open(struct inode *, struct file *); 3283 extern int dcache_dir_close(struct inode *, struct file *); 3284 extern loff_t dcache_dir_lseek(struct file *, loff_t, int); 3285 extern int dcache_readdir(struct file *, struct dir_context *); 3286 extern int simple_setattr(struct mnt_idmap *, struct dentry *, 3287 struct iattr *); 3288 extern int simple_getattr(struct mnt_idmap *, const struct path *, 3289 struct kstat *, u32, unsigned int); 3290 extern int simple_statfs(struct dentry *, struct kstatfs *); 3291 extern int simple_open(struct inode *inode, struct file *file); 3292 extern int simple_link(struct dentry *, struct inode *, struct dentry *); 3293 extern int simple_unlink(struct inode *, struct dentry *); 3294 extern int simple_rmdir(struct inode *, struct dentry *); 3295 void simple_rename_timestamp(struct inode *old_dir, struct dentry *old_dentry, 3296 struct inode *new_dir, struct dentry *new_dentry); 3297 extern int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry, 3298 struct inode *new_dir, struct dentry *new_dentry); 3299 extern int simple_rename(struct mnt_idmap *, struct inode *, 3300 struct dentry *, struct inode *, struct dentry *, 3301 unsigned int); 3302 extern void simple_recursive_removal(struct dentry *, 3303 void (*callback)(struct dentry *)); 3304 extern int noop_fsync(struct file *, loff_t, loff_t, int); 3305 extern ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter); 3306 extern int simple_empty(struct dentry *); 3307 extern int simple_write_begin(struct file *file, struct address_space *mapping, 3308 loff_t pos, unsigned len, 3309 struct page **pagep, void **fsdata); 3310 extern const struct address_space_operations ram_aops; 3311 extern int always_delete_dentry(const struct dentry *); 3312 extern struct inode *alloc_anon_inode(struct super_block *); 3313 extern int simple_nosetlease(struct file *, int, struct file_lease **, void **); 3314 extern const struct dentry_operations simple_dentry_operations; 3315 3316 extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags); 3317 extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *); 3318 extern const struct file_operations simple_dir_operations; 3319 extern const struct inode_operations simple_dir_inode_operations; 3320 extern void make_empty_dir_inode(struct inode *inode); 3321 extern bool is_empty_dir_inode(struct inode *inode); 3322 struct tree_descr { const char *name; const struct file_operations *ops; int mode; }; 3323 struct dentry *d_alloc_name(struct dentry *, const char *); 3324 extern int simple_fill_super(struct super_block *, unsigned long, 3325 const struct tree_descr *); 3326 extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count); 3327 extern void simple_release_fs(struct vfsmount **mount, int *count); 3328 3329 extern ssize_t simple_read_from_buffer(void __user *to, size_t count, 3330 loff_t *ppos, const void *from, size_t available); 3331 extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos, 3332 const void __user *from, size_t count); 3333 3334 struct offset_ctx { 3335 struct maple_tree mt; 3336 unsigned long next_offset; 3337 }; 3338 3339 void simple_offset_init(struct offset_ctx *octx); 3340 int simple_offset_add(struct offset_ctx *octx, struct dentry *dentry); 3341 void simple_offset_remove(struct offset_ctx *octx, struct dentry *dentry); 3342 int simple_offset_empty(struct dentry *dentry); 3343 int simple_offset_rename_exchange(struct inode *old_dir, 3344 struct dentry *old_dentry, 3345 struct inode *new_dir, 3346 struct dentry *new_dentry); 3347 void simple_offset_destroy(struct offset_ctx *octx); 3348 3349 extern const struct file_operations simple_offset_dir_operations; 3350 3351 extern int __generic_file_fsync(struct file *, loff_t, loff_t, int); 3352 extern int generic_file_fsync(struct file *, loff_t, loff_t, int); 3353 3354 extern int generic_check_addressable(unsigned, u64); 3355 3356 extern void generic_set_sb_d_ops(struct super_block *sb); 3357 3358 static inline bool sb_has_encoding(const struct super_block *sb) 3359 { 3360 #if IS_ENABLED(CONFIG_UNICODE) 3361 return !!sb->s_encoding; 3362 #else 3363 return false; 3364 #endif 3365 } 3366 3367 int may_setattr(struct mnt_idmap *idmap, struct inode *inode, 3368 unsigned int ia_valid); 3369 int setattr_prepare(struct mnt_idmap *, struct dentry *, struct iattr *); 3370 extern int inode_newsize_ok(const struct inode *, loff_t offset); 3371 void setattr_copy(struct mnt_idmap *, struct inode *inode, 3372 const struct iattr *attr); 3373 3374 extern int file_update_time(struct file *file); 3375 3376 static inline bool vma_is_dax(const struct vm_area_struct *vma) 3377 { 3378 return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host); 3379 } 3380 3381 static inline bool vma_is_fsdax(struct vm_area_struct *vma) 3382 { 3383 struct inode *inode; 3384 3385 if (!IS_ENABLED(CONFIG_FS_DAX) || !vma->vm_file) 3386 return false; 3387 if (!vma_is_dax(vma)) 3388 return false; 3389 inode = file_inode(vma->vm_file); 3390 if (S_ISCHR(inode->i_mode)) 3391 return false; /* device-dax */ 3392 return true; 3393 } 3394 3395 static inline int iocb_flags(struct file *file) 3396 { 3397 int res = 0; 3398 if (file->f_flags & O_APPEND) 3399 res |= IOCB_APPEND; 3400 if (file->f_flags & O_DIRECT) 3401 res |= IOCB_DIRECT; 3402 if (file->f_flags & O_DSYNC) 3403 res |= IOCB_DSYNC; 3404 if (file->f_flags & __O_SYNC) 3405 res |= IOCB_SYNC; 3406 return res; 3407 } 3408 3409 static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags) 3410 { 3411 int kiocb_flags = 0; 3412 3413 /* make sure there's no overlap between RWF and private IOCB flags */ 3414 BUILD_BUG_ON((__force int) RWF_SUPPORTED & IOCB_EVENTFD); 3415 3416 if (!flags) 3417 return 0; 3418 if (unlikely(flags & ~RWF_SUPPORTED)) 3419 return -EOPNOTSUPP; 3420 if (unlikely((flags & RWF_APPEND) && (flags & RWF_NOAPPEND))) 3421 return -EINVAL; 3422 3423 if (flags & RWF_NOWAIT) { 3424 if (!(ki->ki_filp->f_mode & FMODE_NOWAIT)) 3425 return -EOPNOTSUPP; 3426 kiocb_flags |= IOCB_NOIO; 3427 } 3428 kiocb_flags |= (__force int) (flags & RWF_SUPPORTED); 3429 if (flags & RWF_SYNC) 3430 kiocb_flags |= IOCB_DSYNC; 3431 3432 if ((flags & RWF_NOAPPEND) && (ki->ki_flags & IOCB_APPEND)) { 3433 if (IS_APPEND(file_inode(ki->ki_filp))) 3434 return -EPERM; 3435 ki->ki_flags &= ~IOCB_APPEND; 3436 } 3437 3438 ki->ki_flags |= kiocb_flags; 3439 return 0; 3440 } 3441 3442 static inline ino_t parent_ino(struct dentry *dentry) 3443 { 3444 ino_t res; 3445 3446 /* 3447 * Don't strictly need d_lock here? If the parent ino could change 3448 * then surely we'd have a deeper race in the caller? 3449 */ 3450 spin_lock(&dentry->d_lock); 3451 res = dentry->d_parent->d_inode->i_ino; 3452 spin_unlock(&dentry->d_lock); 3453 return res; 3454 } 3455 3456 /* Transaction based IO helpers */ 3457 3458 /* 3459 * An argresp is stored in an allocated page and holds the 3460 * size of the argument or response, along with its content 3461 */ 3462 struct simple_transaction_argresp { 3463 ssize_t size; 3464 char data[]; 3465 }; 3466 3467 #define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp)) 3468 3469 char *simple_transaction_get(struct file *file, const char __user *buf, 3470 size_t size); 3471 ssize_t simple_transaction_read(struct file *file, char __user *buf, 3472 size_t size, loff_t *pos); 3473 int simple_transaction_release(struct inode *inode, struct file *file); 3474 3475 void simple_transaction_set(struct file *file, size_t n); 3476 3477 /* 3478 * simple attribute files 3479 * 3480 * These attributes behave similar to those in sysfs: 3481 * 3482 * Writing to an attribute immediately sets a value, an open file can be 3483 * written to multiple times. 3484 * 3485 * Reading from an attribute creates a buffer from the value that might get 3486 * read with multiple read calls. When the attribute has been read 3487 * completely, no further read calls are possible until the file is opened 3488 * again. 3489 * 3490 * All attributes contain a text representation of a numeric value 3491 * that are accessed with the get() and set() functions. 3492 */ 3493 #define DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, __is_signed) \ 3494 static int __fops ## _open(struct inode *inode, struct file *file) \ 3495 { \ 3496 __simple_attr_check_format(__fmt, 0ull); \ 3497 return simple_attr_open(inode, file, __get, __set, __fmt); \ 3498 } \ 3499 static const struct file_operations __fops = { \ 3500 .owner = THIS_MODULE, \ 3501 .open = __fops ## _open, \ 3502 .release = simple_attr_release, \ 3503 .read = simple_attr_read, \ 3504 .write = (__is_signed) ? simple_attr_write_signed : simple_attr_write, \ 3505 .llseek = generic_file_llseek, \ 3506 } 3507 3508 #define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt) \ 3509 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, false) 3510 3511 #define DEFINE_SIMPLE_ATTRIBUTE_SIGNED(__fops, __get, __set, __fmt) \ 3512 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, true) 3513 3514 static inline __printf(1, 2) 3515 void __simple_attr_check_format(const char *fmt, ...) 3516 { 3517 /* don't do anything, just let the compiler check the arguments; */ 3518 } 3519 3520 int simple_attr_open(struct inode *inode, struct file *file, 3521 int (*get)(void *, u64 *), int (*set)(void *, u64), 3522 const char *fmt); 3523 int simple_attr_release(struct inode *inode, struct file *file); 3524 ssize_t simple_attr_read(struct file *file, char __user *buf, 3525 size_t len, loff_t *ppos); 3526 ssize_t simple_attr_write(struct file *file, const char __user *buf, 3527 size_t len, loff_t *ppos); 3528 ssize_t simple_attr_write_signed(struct file *file, const char __user *buf, 3529 size_t len, loff_t *ppos); 3530 3531 struct ctl_table; 3532 int __init list_bdev_fs_names(char *buf, size_t size); 3533 3534 #define __FMODE_EXEC ((__force int) FMODE_EXEC) 3535 #define __FMODE_NONOTIFY ((__force int) FMODE_NONOTIFY) 3536 3537 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE]) 3538 #define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \ 3539 (flag & __FMODE_NONOTIFY))) 3540 3541 static inline bool is_sxid(umode_t mode) 3542 { 3543 return mode & (S_ISUID | S_ISGID); 3544 } 3545 3546 static inline int check_sticky(struct mnt_idmap *idmap, 3547 struct inode *dir, struct inode *inode) 3548 { 3549 if (!(dir->i_mode & S_ISVTX)) 3550 return 0; 3551 3552 return __check_sticky(idmap, dir, inode); 3553 } 3554 3555 static inline void inode_has_no_xattr(struct inode *inode) 3556 { 3557 if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC)) 3558 inode->i_flags |= S_NOSEC; 3559 } 3560 3561 static inline bool is_root_inode(struct inode *inode) 3562 { 3563 return inode == inode->i_sb->s_root->d_inode; 3564 } 3565 3566 static inline bool dir_emit(struct dir_context *ctx, 3567 const char *name, int namelen, 3568 u64 ino, unsigned type) 3569 { 3570 return ctx->actor(ctx, name, namelen, ctx->pos, ino, type); 3571 } 3572 static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx) 3573 { 3574 return ctx->actor(ctx, ".", 1, ctx->pos, 3575 file->f_path.dentry->d_inode->i_ino, DT_DIR); 3576 } 3577 static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx) 3578 { 3579 return ctx->actor(ctx, "..", 2, ctx->pos, 3580 parent_ino(file->f_path.dentry), DT_DIR); 3581 } 3582 static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx) 3583 { 3584 if (ctx->pos == 0) { 3585 if (!dir_emit_dot(file, ctx)) 3586 return false; 3587 ctx->pos = 1; 3588 } 3589 if (ctx->pos == 1) { 3590 if (!dir_emit_dotdot(file, ctx)) 3591 return false; 3592 ctx->pos = 2; 3593 } 3594 return true; 3595 } 3596 static inline bool dir_relax(struct inode *inode) 3597 { 3598 inode_unlock(inode); 3599 inode_lock(inode); 3600 return !IS_DEADDIR(inode); 3601 } 3602 3603 static inline bool dir_relax_shared(struct inode *inode) 3604 { 3605 inode_unlock_shared(inode); 3606 inode_lock_shared(inode); 3607 return !IS_DEADDIR(inode); 3608 } 3609 3610 extern bool path_noexec(const struct path *path); 3611 extern void inode_nohighmem(struct inode *inode); 3612 3613 /* mm/fadvise.c */ 3614 extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len, 3615 int advice); 3616 extern int generic_fadvise(struct file *file, loff_t offset, loff_t len, 3617 int advice); 3618 3619 #endif /* _LINUX_FS_H */ 3620