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