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 45 #include <asm/byteorder.h> 46 #include <uapi/linux/fs.h> 47 48 struct backing_dev_info; 49 struct bdi_writeback; 50 struct bio; 51 struct export_operations; 52 struct fiemap_extent_info; 53 struct hd_geometry; 54 struct iovec; 55 struct kiocb; 56 struct kobject; 57 struct pipe_inode_info; 58 struct poll_table_struct; 59 struct kstatfs; 60 struct vm_area_struct; 61 struct vfsmount; 62 struct cred; 63 struct swap_info_struct; 64 struct seq_file; 65 struct workqueue_struct; 66 struct iov_iter; 67 struct fscrypt_info; 68 struct fscrypt_operations; 69 struct fsverity_info; 70 struct fsverity_operations; 71 struct fs_context; 72 struct fs_parameter_spec; 73 struct fileattr; 74 75 extern void __init inode_init(void); 76 extern void __init inode_init_early(void); 77 extern void __init files_init(void); 78 extern void __init files_maxfiles_init(void); 79 80 extern struct files_stat_struct files_stat; 81 extern unsigned long get_max_files(void); 82 extern unsigned int sysctl_nr_open; 83 extern struct inodes_stat_t inodes_stat; 84 extern int leases_enable, lease_break_time; 85 extern int sysctl_protected_symlinks; 86 extern int sysctl_protected_hardlinks; 87 extern int sysctl_protected_fifos; 88 extern int sysctl_protected_regular; 89 90 typedef __kernel_rwf_t rwf_t; 91 92 struct buffer_head; 93 typedef int (get_block_t)(struct inode *inode, sector_t iblock, 94 struct buffer_head *bh_result, int create); 95 typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset, 96 ssize_t bytes, void *private); 97 98 #define MAY_EXEC 0x00000001 99 #define MAY_WRITE 0x00000002 100 #define MAY_READ 0x00000004 101 #define MAY_APPEND 0x00000008 102 #define MAY_ACCESS 0x00000010 103 #define MAY_OPEN 0x00000020 104 #define MAY_CHDIR 0x00000040 105 /* called from RCU mode, don't block */ 106 #define MAY_NOT_BLOCK 0x00000080 107 108 /* 109 * flags in file.f_mode. Note that FMODE_READ and FMODE_WRITE must correspond 110 * to O_WRONLY and O_RDWR via the strange trick in do_dentry_open() 111 */ 112 113 /* file is open for reading */ 114 #define FMODE_READ ((__force fmode_t)0x1) 115 /* file is open for writing */ 116 #define FMODE_WRITE ((__force fmode_t)0x2) 117 /* file is seekable */ 118 #define FMODE_LSEEK ((__force fmode_t)0x4) 119 /* file can be accessed using pread */ 120 #define FMODE_PREAD ((__force fmode_t)0x8) 121 /* file can be accessed using pwrite */ 122 #define FMODE_PWRITE ((__force fmode_t)0x10) 123 /* File is opened for execution with sys_execve / sys_uselib */ 124 #define FMODE_EXEC ((__force fmode_t)0x20) 125 /* File is opened with O_NDELAY (only set for block devices) */ 126 #define FMODE_NDELAY ((__force fmode_t)0x40) 127 /* File is opened with O_EXCL (only set for block devices) */ 128 #define FMODE_EXCL ((__force fmode_t)0x80) 129 /* File is opened using open(.., 3, ..) and is writeable only for ioctls 130 (specialy hack for floppy.c) */ 131 #define FMODE_WRITE_IOCTL ((__force fmode_t)0x100) 132 /* 32bit hashes as llseek() offset (for directories) */ 133 #define FMODE_32BITHASH ((__force fmode_t)0x200) 134 /* 64bit hashes as llseek() offset (for directories) */ 135 #define FMODE_64BITHASH ((__force fmode_t)0x400) 136 137 /* 138 * Don't update ctime and mtime. 139 * 140 * Currently a special hack for the XFS open_by_handle ioctl, but we'll 141 * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon. 142 */ 143 #define FMODE_NOCMTIME ((__force fmode_t)0x800) 144 145 /* Expect random access pattern */ 146 #define FMODE_RANDOM ((__force fmode_t)0x1000) 147 148 /* File is huge (eg. /dev/mem): treat loff_t as unsigned */ 149 #define FMODE_UNSIGNED_OFFSET ((__force fmode_t)0x2000) 150 151 /* File is opened with O_PATH; almost nothing can be done with it */ 152 #define FMODE_PATH ((__force fmode_t)0x4000) 153 154 /* File needs atomic accesses to f_pos */ 155 #define FMODE_ATOMIC_POS ((__force fmode_t)0x8000) 156 /* Write access to underlying fs */ 157 #define FMODE_WRITER ((__force fmode_t)0x10000) 158 /* Has read method(s) */ 159 #define FMODE_CAN_READ ((__force fmode_t)0x20000) 160 /* Has write method(s) */ 161 #define FMODE_CAN_WRITE ((__force fmode_t)0x40000) 162 163 #define FMODE_OPENED ((__force fmode_t)0x80000) 164 #define FMODE_CREATED ((__force fmode_t)0x100000) 165 166 /* File is stream-like */ 167 #define FMODE_STREAM ((__force fmode_t)0x200000) 168 169 /* File was opened by fanotify and shouldn't generate fanotify events */ 170 #define FMODE_NONOTIFY ((__force fmode_t)0x4000000) 171 172 /* File is capable of returning -EAGAIN if I/O will block */ 173 #define FMODE_NOWAIT ((__force fmode_t)0x8000000) 174 175 /* File represents mount that needs unmounting */ 176 #define FMODE_NEED_UNMOUNT ((__force fmode_t)0x10000000) 177 178 /* File does not contribute to nr_files count */ 179 #define FMODE_NOACCOUNT ((__force fmode_t)0x20000000) 180 181 /* File supports async buffered reads */ 182 #define FMODE_BUF_RASYNC ((__force fmode_t)0x40000000) 183 184 /* 185 * Attribute flags. These should be or-ed together to figure out what 186 * has been changed! 187 */ 188 #define ATTR_MODE (1 << 0) 189 #define ATTR_UID (1 << 1) 190 #define ATTR_GID (1 << 2) 191 #define ATTR_SIZE (1 << 3) 192 #define ATTR_ATIME (1 << 4) 193 #define ATTR_MTIME (1 << 5) 194 #define ATTR_CTIME (1 << 6) 195 #define ATTR_ATIME_SET (1 << 7) 196 #define ATTR_MTIME_SET (1 << 8) 197 #define ATTR_FORCE (1 << 9) /* Not a change, but a change it */ 198 #define ATTR_KILL_SUID (1 << 11) 199 #define ATTR_KILL_SGID (1 << 12) 200 #define ATTR_FILE (1 << 13) 201 #define ATTR_KILL_PRIV (1 << 14) 202 #define ATTR_OPEN (1 << 15) /* Truncating from open(O_TRUNC) */ 203 #define ATTR_TIMES_SET (1 << 16) 204 #define ATTR_TOUCH (1 << 17) 205 206 /* 207 * Whiteout is represented by a char device. The following constants define the 208 * mode and device number to use. 209 */ 210 #define WHITEOUT_MODE 0 211 #define WHITEOUT_DEV 0 212 213 /* 214 * This is the Inode Attributes structure, used for notify_change(). It 215 * uses the above definitions as flags, to know which values have changed. 216 * Also, in this manner, a Filesystem can look at only the values it cares 217 * about. Basically, these are the attributes that the VFS layer can 218 * request to change from the FS layer. 219 * 220 * Derek Atkins <warlord@MIT.EDU> 94-10-20 221 */ 222 struct iattr { 223 unsigned int ia_valid; 224 umode_t ia_mode; 225 kuid_t ia_uid; 226 kgid_t ia_gid; 227 loff_t ia_size; 228 struct timespec64 ia_atime; 229 struct timespec64 ia_mtime; 230 struct timespec64 ia_ctime; 231 232 /* 233 * Not an attribute, but an auxiliary info for filesystems wanting to 234 * implement an ftruncate() like method. NOTE: filesystem should 235 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL). 236 */ 237 struct file *ia_file; 238 }; 239 240 /* 241 * Includes for diskquotas. 242 */ 243 #include <linux/quota.h> 244 245 /* 246 * Maximum number of layers of fs stack. Needs to be limited to 247 * prevent kernel stack overflow 248 */ 249 #define FILESYSTEM_MAX_STACK_DEPTH 2 250 251 /** 252 * enum positive_aop_returns - aop return codes with specific semantics 253 * 254 * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has 255 * completed, that the page is still locked, and 256 * should be considered active. The VM uses this hint 257 * to return the page to the active list -- it won't 258 * be a candidate for writeback again in the near 259 * future. Other callers must be careful to unlock 260 * the page if they get this return. Returned by 261 * writepage(); 262 * 263 * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has 264 * unlocked it and the page might have been truncated. 265 * The caller should back up to acquiring a new page and 266 * trying again. The aop will be taking reasonable 267 * precautions not to livelock. If the caller held a page 268 * reference, it should drop it before retrying. Returned 269 * by readpage(). 270 * 271 * address_space_operation functions return these large constants to indicate 272 * special semantics to the caller. These are much larger than the bytes in a 273 * page to allow for functions that return the number of bytes operated on in a 274 * given page. 275 */ 276 277 enum positive_aop_returns { 278 AOP_WRITEPAGE_ACTIVATE = 0x80000, 279 AOP_TRUNCATED_PAGE = 0x80001, 280 }; 281 282 #define AOP_FLAG_CONT_EXPAND 0x0001 /* called from cont_expand */ 283 #define AOP_FLAG_NOFS 0x0002 /* used by filesystem to direct 284 * helper code (eg buffer layer) 285 * to clear GFP_FS from alloc */ 286 287 /* 288 * oh the beauties of C type declarations. 289 */ 290 struct page; 291 struct address_space; 292 struct writeback_control; 293 struct readahead_control; 294 295 /* 296 * Write life time hint values. 297 * Stored in struct inode as u8. 298 */ 299 enum rw_hint { 300 WRITE_LIFE_NOT_SET = 0, 301 WRITE_LIFE_NONE = RWH_WRITE_LIFE_NONE, 302 WRITE_LIFE_SHORT = RWH_WRITE_LIFE_SHORT, 303 WRITE_LIFE_MEDIUM = RWH_WRITE_LIFE_MEDIUM, 304 WRITE_LIFE_LONG = RWH_WRITE_LIFE_LONG, 305 WRITE_LIFE_EXTREME = RWH_WRITE_LIFE_EXTREME, 306 }; 307 308 /* Match RWF_* bits to IOCB bits */ 309 #define IOCB_HIPRI (__force int) RWF_HIPRI 310 #define IOCB_DSYNC (__force int) RWF_DSYNC 311 #define IOCB_SYNC (__force int) RWF_SYNC 312 #define IOCB_NOWAIT (__force int) RWF_NOWAIT 313 #define IOCB_APPEND (__force int) RWF_APPEND 314 315 /* non-RWF related bits - start at 16 */ 316 #define IOCB_EVENTFD (1 << 16) 317 #define IOCB_DIRECT (1 << 17) 318 #define IOCB_WRITE (1 << 18) 319 /* iocb->ki_waitq is valid */ 320 #define IOCB_WAITQ (1 << 19) 321 #define IOCB_NOIO (1 << 20) 322 /* can use bio alloc cache */ 323 #define IOCB_ALLOC_CACHE (1 << 21) 324 325 struct kiocb { 326 struct file *ki_filp; 327 328 /* The 'ki_filp' pointer is shared in a union for aio */ 329 randomized_struct_fields_start 330 331 loff_t ki_pos; 332 void (*ki_complete)(struct kiocb *iocb, long ret, long ret2); 333 void *private; 334 int ki_flags; 335 u16 ki_hint; 336 u16 ki_ioprio; /* See linux/ioprio.h */ 337 union { 338 unsigned int ki_cookie; /* for ->iopoll */ 339 struct wait_page_queue *ki_waitq; /* for async buffered IO */ 340 }; 341 342 randomized_struct_fields_end 343 }; 344 345 static inline bool is_sync_kiocb(struct kiocb *kiocb) 346 { 347 return kiocb->ki_complete == NULL; 348 } 349 350 /* 351 * "descriptor" for what we're up to with a read. 352 * This allows us to use the same read code yet 353 * have multiple different users of the data that 354 * we read from a file. 355 * 356 * The simplest case just copies the data to user 357 * mode. 358 */ 359 typedef struct { 360 size_t written; 361 size_t count; 362 union { 363 char __user *buf; 364 void *data; 365 } arg; 366 int error; 367 } read_descriptor_t; 368 369 typedef int (*read_actor_t)(read_descriptor_t *, struct page *, 370 unsigned long, unsigned long); 371 372 struct address_space_operations { 373 int (*writepage)(struct page *page, struct writeback_control *wbc); 374 int (*readpage)(struct file *, struct page *); 375 376 /* Write back some dirty pages from this mapping. */ 377 int (*writepages)(struct address_space *, struct writeback_control *); 378 379 /* Set a page dirty. Return true if this dirtied it */ 380 int (*set_page_dirty)(struct page *page); 381 382 /* 383 * Reads in the requested pages. Unlike ->readpage(), this is 384 * PURELY used for read-ahead!. 385 */ 386 int (*readpages)(struct file *filp, struct address_space *mapping, 387 struct list_head *pages, unsigned nr_pages); 388 void (*readahead)(struct readahead_control *); 389 390 int (*write_begin)(struct file *, struct address_space *mapping, 391 loff_t pos, unsigned len, unsigned flags, 392 struct page **pagep, void **fsdata); 393 int (*write_end)(struct file *, struct address_space *mapping, 394 loff_t pos, unsigned len, unsigned copied, 395 struct page *page, void *fsdata); 396 397 /* Unfortunately this kludge is needed for FIBMAP. Don't use it */ 398 sector_t (*bmap)(struct address_space *, sector_t); 399 void (*invalidatepage) (struct page *, unsigned int, unsigned int); 400 int (*releasepage) (struct page *, gfp_t); 401 void (*freepage)(struct page *); 402 ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter); 403 /* 404 * migrate the contents of a page to the specified target. If 405 * migrate_mode is MIGRATE_ASYNC, it must not block. 406 */ 407 int (*migratepage) (struct address_space *, 408 struct page *, struct page *, enum migrate_mode); 409 bool (*isolate_page)(struct page *, isolate_mode_t); 410 void (*putback_page)(struct page *); 411 int (*launder_page) (struct page *); 412 int (*is_partially_uptodate) (struct page *, unsigned long, 413 unsigned long); 414 void (*is_dirty_writeback) (struct page *, bool *, bool *); 415 int (*error_remove_page)(struct address_space *, struct page *); 416 417 /* swapfile support */ 418 int (*swap_activate)(struct swap_info_struct *sis, struct file *file, 419 sector_t *span); 420 void (*swap_deactivate)(struct file *file); 421 }; 422 423 extern const struct address_space_operations empty_aops; 424 425 /* 426 * pagecache_write_begin/pagecache_write_end must be used by general code 427 * to write into the pagecache. 428 */ 429 int pagecache_write_begin(struct file *, struct address_space *mapping, 430 loff_t pos, unsigned len, unsigned flags, 431 struct page **pagep, void **fsdata); 432 433 int pagecache_write_end(struct file *, struct address_space *mapping, 434 loff_t pos, unsigned len, unsigned copied, 435 struct page *page, void *fsdata); 436 437 /** 438 * struct address_space - Contents of a cacheable, mappable object. 439 * @host: Owner, either the inode or the block_device. 440 * @i_pages: Cached pages. 441 * @invalidate_lock: Guards coherency between page cache contents and 442 * file offset->disk block mappings in the filesystem during invalidates. 443 * It is also used to block modification of page cache contents through 444 * memory mappings. 445 * @gfp_mask: Memory allocation flags to use for allocating pages. 446 * @i_mmap_writable: Number of VM_SHARED mappings. 447 * @nr_thps: Number of THPs in the pagecache (non-shmem only). 448 * @i_mmap: Tree of private and shared mappings. 449 * @i_mmap_rwsem: Protects @i_mmap and @i_mmap_writable. 450 * @nrpages: Number of page entries, protected by the i_pages lock. 451 * @writeback_index: Writeback starts here. 452 * @a_ops: Methods. 453 * @flags: Error bits and flags (AS_*). 454 * @wb_err: The most recent error which has occurred. 455 * @private_lock: For use by the owner of the address_space. 456 * @private_list: For use by the owner of the address_space. 457 * @private_data: For use by the owner of the address_space. 458 */ 459 struct address_space { 460 struct inode *host; 461 struct xarray i_pages; 462 struct rw_semaphore invalidate_lock; 463 gfp_t gfp_mask; 464 atomic_t i_mmap_writable; 465 #ifdef CONFIG_READ_ONLY_THP_FOR_FS 466 /* number of thp, only for non-shmem files */ 467 atomic_t nr_thps; 468 #endif 469 struct rb_root_cached i_mmap; 470 struct rw_semaphore i_mmap_rwsem; 471 unsigned long nrpages; 472 pgoff_t writeback_index; 473 const struct address_space_operations *a_ops; 474 unsigned long flags; 475 errseq_t wb_err; 476 spinlock_t private_lock; 477 struct list_head private_list; 478 void *private_data; 479 } __attribute__((aligned(sizeof(long)))) __randomize_layout; 480 /* 481 * On most architectures that alignment is already the case; but 482 * must be enforced here for CRIS, to let the least significant bit 483 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON. 484 */ 485 486 /* XArray tags, for tagging dirty and writeback pages in the pagecache. */ 487 #define PAGECACHE_TAG_DIRTY XA_MARK_0 488 #define PAGECACHE_TAG_WRITEBACK XA_MARK_1 489 #define PAGECACHE_TAG_TOWRITE XA_MARK_2 490 491 /* 492 * Returns true if any of the pages in the mapping are marked with the tag. 493 */ 494 static inline bool mapping_tagged(struct address_space *mapping, xa_mark_t tag) 495 { 496 return xa_marked(&mapping->i_pages, tag); 497 } 498 499 static inline void i_mmap_lock_write(struct address_space *mapping) 500 { 501 down_write(&mapping->i_mmap_rwsem); 502 } 503 504 static inline int i_mmap_trylock_write(struct address_space *mapping) 505 { 506 return down_write_trylock(&mapping->i_mmap_rwsem); 507 } 508 509 static inline void i_mmap_unlock_write(struct address_space *mapping) 510 { 511 up_write(&mapping->i_mmap_rwsem); 512 } 513 514 static inline void i_mmap_lock_read(struct address_space *mapping) 515 { 516 down_read(&mapping->i_mmap_rwsem); 517 } 518 519 static inline void i_mmap_unlock_read(struct address_space *mapping) 520 { 521 up_read(&mapping->i_mmap_rwsem); 522 } 523 524 static inline void i_mmap_assert_locked(struct address_space *mapping) 525 { 526 lockdep_assert_held(&mapping->i_mmap_rwsem); 527 } 528 529 static inline void i_mmap_assert_write_locked(struct address_space *mapping) 530 { 531 lockdep_assert_held_write(&mapping->i_mmap_rwsem); 532 } 533 534 /* 535 * Might pages of this file be mapped into userspace? 536 */ 537 static inline int mapping_mapped(struct address_space *mapping) 538 { 539 return !RB_EMPTY_ROOT(&mapping->i_mmap.rb_root); 540 } 541 542 /* 543 * Might pages of this file have been modified in userspace? 544 * Note that i_mmap_writable counts all VM_SHARED vmas: do_mmap 545 * marks vma as VM_SHARED if it is shared, and the file was opened for 546 * writing i.e. vma may be mprotected writable even if now readonly. 547 * 548 * If i_mmap_writable is negative, no new writable mappings are allowed. You 549 * can only deny writable mappings, if none exists right now. 550 */ 551 static inline int mapping_writably_mapped(struct address_space *mapping) 552 { 553 return atomic_read(&mapping->i_mmap_writable) > 0; 554 } 555 556 static inline int mapping_map_writable(struct address_space *mapping) 557 { 558 return atomic_inc_unless_negative(&mapping->i_mmap_writable) ? 559 0 : -EPERM; 560 } 561 562 static inline void mapping_unmap_writable(struct address_space *mapping) 563 { 564 atomic_dec(&mapping->i_mmap_writable); 565 } 566 567 static inline int mapping_deny_writable(struct address_space *mapping) 568 { 569 return atomic_dec_unless_positive(&mapping->i_mmap_writable) ? 570 0 : -EBUSY; 571 } 572 573 static inline void mapping_allow_writable(struct address_space *mapping) 574 { 575 atomic_inc(&mapping->i_mmap_writable); 576 } 577 578 /* 579 * Use sequence counter to get consistent i_size on 32-bit processors. 580 */ 581 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 582 #include <linux/seqlock.h> 583 #define __NEED_I_SIZE_ORDERED 584 #define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount) 585 #else 586 #define i_size_ordered_init(inode) do { } while (0) 587 #endif 588 589 struct posix_acl; 590 #define ACL_NOT_CACHED ((void *)(-1)) 591 /* 592 * ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to 593 * cache the ACL. This also means that ->get_acl() can be called in RCU mode 594 * with the LOOKUP_RCU flag. 595 */ 596 #define ACL_DONT_CACHE ((void *)(-3)) 597 598 static inline struct posix_acl * 599 uncached_acl_sentinel(struct task_struct *task) 600 { 601 return (void *)task + 1; 602 } 603 604 static inline bool 605 is_uncached_acl(struct posix_acl *acl) 606 { 607 return (long)acl & 1; 608 } 609 610 #define IOP_FASTPERM 0x0001 611 #define IOP_LOOKUP 0x0002 612 #define IOP_NOFOLLOW 0x0004 613 #define IOP_XATTR 0x0008 614 #define IOP_DEFAULT_READLINK 0x0010 615 616 struct fsnotify_mark_connector; 617 618 /* 619 * Keep mostly read-only and often accessed (especially for 620 * the RCU path lookup and 'stat' data) fields at the beginning 621 * of the 'struct inode' 622 */ 623 struct inode { 624 umode_t i_mode; 625 unsigned short i_opflags; 626 kuid_t i_uid; 627 kgid_t i_gid; 628 unsigned int i_flags; 629 630 #ifdef CONFIG_FS_POSIX_ACL 631 struct posix_acl *i_acl; 632 struct posix_acl *i_default_acl; 633 #endif 634 635 const struct inode_operations *i_op; 636 struct super_block *i_sb; 637 struct address_space *i_mapping; 638 639 #ifdef CONFIG_SECURITY 640 void *i_security; 641 #endif 642 643 /* Stat data, not accessed from path walking */ 644 unsigned long i_ino; 645 /* 646 * Filesystems may only read i_nlink directly. They shall use the 647 * following functions for modification: 648 * 649 * (set|clear|inc|drop)_nlink 650 * inode_(inc|dec)_link_count 651 */ 652 union { 653 const unsigned int i_nlink; 654 unsigned int __i_nlink; 655 }; 656 dev_t i_rdev; 657 loff_t i_size; 658 struct timespec64 i_atime; 659 struct timespec64 i_mtime; 660 struct timespec64 i_ctime; 661 spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */ 662 unsigned short i_bytes; 663 u8 i_blkbits; 664 u8 i_write_hint; 665 blkcnt_t i_blocks; 666 667 #ifdef __NEED_I_SIZE_ORDERED 668 seqcount_t i_size_seqcount; 669 #endif 670 671 /* Misc */ 672 unsigned long i_state; 673 struct rw_semaphore i_rwsem; 674 675 unsigned long dirtied_when; /* jiffies of first dirtying */ 676 unsigned long dirtied_time_when; 677 678 struct hlist_node i_hash; 679 struct list_head i_io_list; /* backing dev IO list */ 680 #ifdef CONFIG_CGROUP_WRITEBACK 681 struct bdi_writeback *i_wb; /* the associated cgroup wb */ 682 683 /* foreign inode detection, see wbc_detach_inode() */ 684 int i_wb_frn_winner; 685 u16 i_wb_frn_avg_time; 686 u16 i_wb_frn_history; 687 #endif 688 struct list_head i_lru; /* inode LRU list */ 689 struct list_head i_sb_list; 690 struct list_head i_wb_list; /* backing dev writeback list */ 691 union { 692 struct hlist_head i_dentry; 693 struct rcu_head i_rcu; 694 }; 695 atomic64_t i_version; 696 atomic64_t i_sequence; /* see futex */ 697 atomic_t i_count; 698 atomic_t i_dio_count; 699 atomic_t i_writecount; 700 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING) 701 atomic_t i_readcount; /* struct files open RO */ 702 #endif 703 union { 704 const struct file_operations *i_fop; /* former ->i_op->default_file_ops */ 705 void (*free_inode)(struct inode *); 706 }; 707 struct file_lock_context *i_flctx; 708 struct address_space i_data; 709 struct list_head i_devices; 710 union { 711 struct pipe_inode_info *i_pipe; 712 struct cdev *i_cdev; 713 char *i_link; 714 unsigned i_dir_seq; 715 }; 716 717 __u32 i_generation; 718 719 #ifdef CONFIG_FSNOTIFY 720 __u32 i_fsnotify_mask; /* all events this inode cares about */ 721 struct fsnotify_mark_connector __rcu *i_fsnotify_marks; 722 #endif 723 724 #ifdef CONFIG_FS_ENCRYPTION 725 struct fscrypt_info *i_crypt_info; 726 #endif 727 728 #ifdef CONFIG_FS_VERITY 729 struct fsverity_info *i_verity_info; 730 #endif 731 732 void *i_private; /* fs or device private pointer */ 733 } __randomize_layout; 734 735 struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode); 736 737 static inline unsigned int i_blocksize(const struct inode *node) 738 { 739 return (1 << node->i_blkbits); 740 } 741 742 static inline int inode_unhashed(struct inode *inode) 743 { 744 return hlist_unhashed(&inode->i_hash); 745 } 746 747 /* 748 * __mark_inode_dirty expects inodes to be hashed. Since we don't 749 * want special inodes in the fileset inode space, we make them 750 * appear hashed, but do not put on any lists. hlist_del() 751 * will work fine and require no locking. 752 */ 753 static inline void inode_fake_hash(struct inode *inode) 754 { 755 hlist_add_fake(&inode->i_hash); 756 } 757 758 /* 759 * inode->i_mutex nesting subclasses for the lock validator: 760 * 761 * 0: the object of the current VFS operation 762 * 1: parent 763 * 2: child/target 764 * 3: xattr 765 * 4: second non-directory 766 * 5: second parent (when locking independent directories in rename) 767 * 768 * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two 769 * non-directories at once. 770 * 771 * The locking order between these classes is 772 * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory 773 */ 774 enum inode_i_mutex_lock_class 775 { 776 I_MUTEX_NORMAL, 777 I_MUTEX_PARENT, 778 I_MUTEX_CHILD, 779 I_MUTEX_XATTR, 780 I_MUTEX_NONDIR2, 781 I_MUTEX_PARENT2, 782 }; 783 784 static inline void inode_lock(struct inode *inode) 785 { 786 down_write(&inode->i_rwsem); 787 } 788 789 static inline void inode_unlock(struct inode *inode) 790 { 791 up_write(&inode->i_rwsem); 792 } 793 794 static inline void inode_lock_shared(struct inode *inode) 795 { 796 down_read(&inode->i_rwsem); 797 } 798 799 static inline void inode_unlock_shared(struct inode *inode) 800 { 801 up_read(&inode->i_rwsem); 802 } 803 804 static inline int inode_trylock(struct inode *inode) 805 { 806 return down_write_trylock(&inode->i_rwsem); 807 } 808 809 static inline int inode_trylock_shared(struct inode *inode) 810 { 811 return down_read_trylock(&inode->i_rwsem); 812 } 813 814 static inline int inode_is_locked(struct inode *inode) 815 { 816 return rwsem_is_locked(&inode->i_rwsem); 817 } 818 819 static inline void inode_lock_nested(struct inode *inode, unsigned subclass) 820 { 821 down_write_nested(&inode->i_rwsem, subclass); 822 } 823 824 static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass) 825 { 826 down_read_nested(&inode->i_rwsem, subclass); 827 } 828 829 static inline void filemap_invalidate_lock(struct address_space *mapping) 830 { 831 down_write(&mapping->invalidate_lock); 832 } 833 834 static inline void filemap_invalidate_unlock(struct address_space *mapping) 835 { 836 up_write(&mapping->invalidate_lock); 837 } 838 839 static inline void filemap_invalidate_lock_shared(struct address_space *mapping) 840 { 841 down_read(&mapping->invalidate_lock); 842 } 843 844 static inline int filemap_invalidate_trylock_shared( 845 struct address_space *mapping) 846 { 847 return down_read_trylock(&mapping->invalidate_lock); 848 } 849 850 static inline void filemap_invalidate_unlock_shared( 851 struct address_space *mapping) 852 { 853 up_read(&mapping->invalidate_lock); 854 } 855 856 void lock_two_nondirectories(struct inode *, struct inode*); 857 void unlock_two_nondirectories(struct inode *, struct inode*); 858 859 void filemap_invalidate_lock_two(struct address_space *mapping1, 860 struct address_space *mapping2); 861 void filemap_invalidate_unlock_two(struct address_space *mapping1, 862 struct address_space *mapping2); 863 864 865 /* 866 * NOTE: in a 32bit arch with a preemptable kernel and 867 * an UP compile the i_size_read/write must be atomic 868 * with respect to the local cpu (unlike with preempt disabled), 869 * but they don't need to be atomic with respect to other cpus like in 870 * true SMP (so they need either to either locally disable irq around 871 * the read or for example on x86 they can be still implemented as a 872 * cmpxchg8b without the need of the lock prefix). For SMP compiles 873 * and 64bit archs it makes no difference if preempt is enabled or not. 874 */ 875 static inline loff_t i_size_read(const struct inode *inode) 876 { 877 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 878 loff_t i_size; 879 unsigned int seq; 880 881 do { 882 seq = read_seqcount_begin(&inode->i_size_seqcount); 883 i_size = inode->i_size; 884 } while (read_seqcount_retry(&inode->i_size_seqcount, seq)); 885 return i_size; 886 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION) 887 loff_t i_size; 888 889 preempt_disable(); 890 i_size = inode->i_size; 891 preempt_enable(); 892 return i_size; 893 #else 894 return inode->i_size; 895 #endif 896 } 897 898 /* 899 * NOTE: unlike i_size_read(), i_size_write() does need locking around it 900 * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount 901 * can be lost, resulting in subsequent i_size_read() calls spinning forever. 902 */ 903 static inline void i_size_write(struct inode *inode, loff_t i_size) 904 { 905 #if BITS_PER_LONG==32 && defined(CONFIG_SMP) 906 preempt_disable(); 907 write_seqcount_begin(&inode->i_size_seqcount); 908 inode->i_size = i_size; 909 write_seqcount_end(&inode->i_size_seqcount); 910 preempt_enable(); 911 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION) 912 preempt_disable(); 913 inode->i_size = i_size; 914 preempt_enable(); 915 #else 916 inode->i_size = i_size; 917 #endif 918 } 919 920 static inline unsigned iminor(const struct inode *inode) 921 { 922 return MINOR(inode->i_rdev); 923 } 924 925 static inline unsigned imajor(const struct inode *inode) 926 { 927 return MAJOR(inode->i_rdev); 928 } 929 930 struct fown_struct { 931 rwlock_t lock; /* protects pid, uid, euid fields */ 932 struct pid *pid; /* pid or -pgrp where SIGIO should be sent */ 933 enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */ 934 kuid_t uid, euid; /* uid/euid of process setting the owner */ 935 int signum; /* posix.1b rt signal to be delivered on IO */ 936 }; 937 938 /** 939 * struct file_ra_state - Track a file's readahead state. 940 * @start: Where the most recent readahead started. 941 * @size: Number of pages read in the most recent readahead. 942 * @async_size: Start next readahead when this many pages are left. 943 * @ra_pages: Maximum size of a readahead request. 944 * @mmap_miss: How many mmap accesses missed in the page cache. 945 * @prev_pos: The last byte in the most recent read request. 946 */ 947 struct file_ra_state { 948 pgoff_t start; 949 unsigned int size; 950 unsigned int async_size; 951 unsigned int ra_pages; 952 unsigned int mmap_miss; 953 loff_t prev_pos; 954 }; 955 956 /* 957 * Check if @index falls in the readahead windows. 958 */ 959 static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index) 960 { 961 return (index >= ra->start && 962 index < ra->start + ra->size); 963 } 964 965 struct file { 966 union { 967 struct llist_node fu_llist; 968 struct rcu_head fu_rcuhead; 969 } f_u; 970 struct path f_path; 971 struct inode *f_inode; /* cached value */ 972 const struct file_operations *f_op; 973 974 /* 975 * Protects f_ep, f_flags. 976 * Must not be taken from IRQ context. 977 */ 978 spinlock_t f_lock; 979 enum rw_hint f_write_hint; 980 atomic_long_t f_count; 981 unsigned int f_flags; 982 fmode_t f_mode; 983 struct mutex f_pos_lock; 984 loff_t f_pos; 985 struct fown_struct f_owner; 986 const struct cred *f_cred; 987 struct file_ra_state f_ra; 988 989 u64 f_version; 990 #ifdef CONFIG_SECURITY 991 void *f_security; 992 #endif 993 /* needed for tty driver, and maybe others */ 994 void *private_data; 995 996 #ifdef CONFIG_EPOLL 997 /* Used by fs/eventpoll.c to link all the hooks to this file */ 998 struct hlist_head *f_ep; 999 #endif /* #ifdef CONFIG_EPOLL */ 1000 struct address_space *f_mapping; 1001 errseq_t f_wb_err; 1002 errseq_t f_sb_err; /* for syncfs */ 1003 } __randomize_layout 1004 __attribute__((aligned(4))); /* lest something weird decides that 2 is OK */ 1005 1006 struct file_handle { 1007 __u32 handle_bytes; 1008 int handle_type; 1009 /* file identifier */ 1010 unsigned char f_handle[]; 1011 }; 1012 1013 static inline struct file *get_file(struct file *f) 1014 { 1015 atomic_long_inc(&f->f_count); 1016 return f; 1017 } 1018 #define get_file_rcu_many(x, cnt) \ 1019 atomic_long_add_unless(&(x)->f_count, (cnt), 0) 1020 #define get_file_rcu(x) get_file_rcu_many((x), 1) 1021 #define file_count(x) atomic_long_read(&(x)->f_count) 1022 1023 #define MAX_NON_LFS ((1UL<<31) - 1) 1024 1025 /* Page cache limit. The filesystems should put that into their s_maxbytes 1026 limits, otherwise bad things can happen in VM. */ 1027 #if BITS_PER_LONG==32 1028 #define MAX_LFS_FILESIZE ((loff_t)ULONG_MAX << PAGE_SHIFT) 1029 #elif BITS_PER_LONG==64 1030 #define MAX_LFS_FILESIZE ((loff_t)LLONG_MAX) 1031 #endif 1032 1033 #define FL_POSIX 1 1034 #define FL_FLOCK 2 1035 #define FL_DELEG 4 /* NFSv4 delegation */ 1036 #define FL_ACCESS 8 /* not trying to lock, just looking */ 1037 #define FL_EXISTS 16 /* when unlocking, test for existence */ 1038 #define FL_LEASE 32 /* lease held on this file */ 1039 #define FL_CLOSE 64 /* unlock on close */ 1040 #define FL_SLEEP 128 /* A blocking lock */ 1041 #define FL_DOWNGRADE_PENDING 256 /* Lease is being downgraded */ 1042 #define FL_UNLOCK_PENDING 512 /* Lease is being broken */ 1043 #define FL_OFDLCK 1024 /* lock is "owned" by struct file */ 1044 #define FL_LAYOUT 2048 /* outstanding pNFS layout */ 1045 #define FL_RECLAIM 4096 /* reclaiming from a reboot server */ 1046 1047 #define FL_CLOSE_POSIX (FL_POSIX | FL_CLOSE) 1048 1049 /* 1050 * Special return value from posix_lock_file() and vfs_lock_file() for 1051 * asynchronous locking. 1052 */ 1053 #define FILE_LOCK_DEFERRED 1 1054 1055 /* legacy typedef, should eventually be removed */ 1056 typedef void *fl_owner_t; 1057 1058 struct file_lock; 1059 1060 struct file_lock_operations { 1061 void (*fl_copy_lock)(struct file_lock *, struct file_lock *); 1062 void (*fl_release_private)(struct file_lock *); 1063 }; 1064 1065 struct lock_manager_operations { 1066 fl_owner_t (*lm_get_owner)(fl_owner_t); 1067 void (*lm_put_owner)(fl_owner_t); 1068 void (*lm_notify)(struct file_lock *); /* unblock callback */ 1069 int (*lm_grant)(struct file_lock *, int); 1070 bool (*lm_break)(struct file_lock *); 1071 int (*lm_change)(struct file_lock *, int, struct list_head *); 1072 void (*lm_setup)(struct file_lock *, void **); 1073 bool (*lm_breaker_owns_lease)(struct file_lock *); 1074 }; 1075 1076 struct lock_manager { 1077 struct list_head list; 1078 /* 1079 * NFSv4 and up also want opens blocked during the grace period; 1080 * NLM doesn't care: 1081 */ 1082 bool block_opens; 1083 }; 1084 1085 struct net; 1086 void locks_start_grace(struct net *, struct lock_manager *); 1087 void locks_end_grace(struct lock_manager *); 1088 bool locks_in_grace(struct net *); 1089 bool opens_in_grace(struct net *); 1090 1091 /* that will die - we need it for nfs_lock_info */ 1092 #include <linux/nfs_fs_i.h> 1093 1094 /* 1095 * struct file_lock represents a generic "file lock". It's used to represent 1096 * POSIX byte range locks, BSD (flock) locks, and leases. It's important to 1097 * note that the same struct is used to represent both a request for a lock and 1098 * the lock itself, but the same object is never used for both. 1099 * 1100 * FIXME: should we create a separate "struct lock_request" to help distinguish 1101 * these two uses? 1102 * 1103 * The varous i_flctx lists are ordered by: 1104 * 1105 * 1) lock owner 1106 * 2) lock range start 1107 * 3) lock range end 1108 * 1109 * Obviously, the last two criteria only matter for POSIX locks. 1110 */ 1111 struct file_lock { 1112 struct file_lock *fl_blocker; /* The lock, that is blocking us */ 1113 struct list_head fl_list; /* link into file_lock_context */ 1114 struct hlist_node fl_link; /* node in global lists */ 1115 struct list_head fl_blocked_requests; /* list of requests with 1116 * ->fl_blocker pointing here 1117 */ 1118 struct list_head fl_blocked_member; /* node in 1119 * ->fl_blocker->fl_blocked_requests 1120 */ 1121 fl_owner_t fl_owner; 1122 unsigned int fl_flags; 1123 unsigned char fl_type; 1124 unsigned int fl_pid; 1125 int fl_link_cpu; /* what cpu's list is this on? */ 1126 wait_queue_head_t fl_wait; 1127 struct file *fl_file; 1128 loff_t fl_start; 1129 loff_t fl_end; 1130 1131 struct fasync_struct * fl_fasync; /* for lease break notifications */ 1132 /* for lease breaks: */ 1133 unsigned long fl_break_time; 1134 unsigned long fl_downgrade_time; 1135 1136 const struct file_lock_operations *fl_ops; /* Callbacks for filesystems */ 1137 const struct lock_manager_operations *fl_lmops; /* Callbacks for lockmanagers */ 1138 union { 1139 struct nfs_lock_info nfs_fl; 1140 struct nfs4_lock_info nfs4_fl; 1141 struct { 1142 struct list_head link; /* link in AFS vnode's pending_locks list */ 1143 int state; /* state of grant or error if -ve */ 1144 unsigned int debug_id; 1145 } afs; 1146 } fl_u; 1147 } __randomize_layout; 1148 1149 struct file_lock_context { 1150 spinlock_t flc_lock; 1151 struct list_head flc_flock; 1152 struct list_head flc_posix; 1153 struct list_head flc_lease; 1154 }; 1155 1156 /* The following constant reflects the upper bound of the file/locking space */ 1157 #ifndef OFFSET_MAX 1158 #define INT_LIMIT(x) (~((x)1 << (sizeof(x)*8 - 1))) 1159 #define OFFSET_MAX INT_LIMIT(loff_t) 1160 #define OFFT_OFFSET_MAX INT_LIMIT(off_t) 1161 #endif 1162 1163 extern void send_sigio(struct fown_struct *fown, int fd, int band); 1164 1165 #define locks_inode(f) file_inode(f) 1166 1167 #ifdef CONFIG_FILE_LOCKING 1168 extern int fcntl_getlk(struct file *, unsigned int, struct flock *); 1169 extern int fcntl_setlk(unsigned int, struct file *, unsigned int, 1170 struct flock *); 1171 1172 #if BITS_PER_LONG == 32 1173 extern int fcntl_getlk64(struct file *, unsigned int, struct flock64 *); 1174 extern int fcntl_setlk64(unsigned int, struct file *, unsigned int, 1175 struct flock64 *); 1176 #endif 1177 1178 extern int fcntl_setlease(unsigned int fd, struct file *filp, long arg); 1179 extern int fcntl_getlease(struct file *filp); 1180 1181 /* fs/locks.c */ 1182 void locks_free_lock_context(struct inode *inode); 1183 void locks_free_lock(struct file_lock *fl); 1184 extern void locks_init_lock(struct file_lock *); 1185 extern struct file_lock * locks_alloc_lock(void); 1186 extern void locks_copy_lock(struct file_lock *, struct file_lock *); 1187 extern void locks_copy_conflock(struct file_lock *, struct file_lock *); 1188 extern void locks_remove_posix(struct file *, fl_owner_t); 1189 extern void locks_remove_file(struct file *); 1190 extern void locks_release_private(struct file_lock *); 1191 extern void posix_test_lock(struct file *, struct file_lock *); 1192 extern int posix_lock_file(struct file *, struct file_lock *, struct file_lock *); 1193 extern int locks_delete_block(struct file_lock *); 1194 extern int vfs_test_lock(struct file *, struct file_lock *); 1195 extern int vfs_lock_file(struct file *, unsigned int, struct file_lock *, struct file_lock *); 1196 extern int vfs_cancel_lock(struct file *filp, struct file_lock *fl); 1197 extern int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl); 1198 extern int __break_lease(struct inode *inode, unsigned int flags, unsigned int type); 1199 extern void lease_get_mtime(struct inode *, struct timespec64 *time); 1200 extern int generic_setlease(struct file *, long, struct file_lock **, void **priv); 1201 extern int vfs_setlease(struct file *, long, struct file_lock **, void **); 1202 extern int lease_modify(struct file_lock *, int, struct list_head *); 1203 1204 struct notifier_block; 1205 extern int lease_register_notifier(struct notifier_block *); 1206 extern void lease_unregister_notifier(struct notifier_block *); 1207 1208 struct files_struct; 1209 extern void show_fd_locks(struct seq_file *f, 1210 struct file *filp, struct files_struct *files); 1211 #else /* !CONFIG_FILE_LOCKING */ 1212 static inline int fcntl_getlk(struct file *file, unsigned int cmd, 1213 struct flock __user *user) 1214 { 1215 return -EINVAL; 1216 } 1217 1218 static inline int fcntl_setlk(unsigned int fd, struct file *file, 1219 unsigned int cmd, struct flock __user *user) 1220 { 1221 return -EACCES; 1222 } 1223 1224 #if BITS_PER_LONG == 32 1225 static inline int fcntl_getlk64(struct file *file, unsigned int cmd, 1226 struct flock64 __user *user) 1227 { 1228 return -EINVAL; 1229 } 1230 1231 static inline int fcntl_setlk64(unsigned int fd, struct file *file, 1232 unsigned int cmd, struct flock64 __user *user) 1233 { 1234 return -EACCES; 1235 } 1236 #endif 1237 static inline int fcntl_setlease(unsigned int fd, struct file *filp, long arg) 1238 { 1239 return -EINVAL; 1240 } 1241 1242 static inline int fcntl_getlease(struct file *filp) 1243 { 1244 return F_UNLCK; 1245 } 1246 1247 static inline void 1248 locks_free_lock_context(struct inode *inode) 1249 { 1250 } 1251 1252 static inline void locks_init_lock(struct file_lock *fl) 1253 { 1254 return; 1255 } 1256 1257 static inline void locks_copy_conflock(struct file_lock *new, struct file_lock *fl) 1258 { 1259 return; 1260 } 1261 1262 static inline void locks_copy_lock(struct file_lock *new, struct file_lock *fl) 1263 { 1264 return; 1265 } 1266 1267 static inline void locks_remove_posix(struct file *filp, fl_owner_t owner) 1268 { 1269 return; 1270 } 1271 1272 static inline void locks_remove_file(struct file *filp) 1273 { 1274 return; 1275 } 1276 1277 static inline void posix_test_lock(struct file *filp, struct file_lock *fl) 1278 { 1279 return; 1280 } 1281 1282 static inline int posix_lock_file(struct file *filp, struct file_lock *fl, 1283 struct file_lock *conflock) 1284 { 1285 return -ENOLCK; 1286 } 1287 1288 static inline int locks_delete_block(struct file_lock *waiter) 1289 { 1290 return -ENOENT; 1291 } 1292 1293 static inline int vfs_test_lock(struct file *filp, struct file_lock *fl) 1294 { 1295 return 0; 1296 } 1297 1298 static inline int vfs_lock_file(struct file *filp, unsigned int cmd, 1299 struct file_lock *fl, struct file_lock *conf) 1300 { 1301 return -ENOLCK; 1302 } 1303 1304 static inline int vfs_cancel_lock(struct file *filp, struct file_lock *fl) 1305 { 1306 return 0; 1307 } 1308 1309 static inline int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl) 1310 { 1311 return -ENOLCK; 1312 } 1313 1314 static inline int __break_lease(struct inode *inode, unsigned int mode, unsigned int type) 1315 { 1316 return 0; 1317 } 1318 1319 static inline void lease_get_mtime(struct inode *inode, 1320 struct timespec64 *time) 1321 { 1322 return; 1323 } 1324 1325 static inline int generic_setlease(struct file *filp, long arg, 1326 struct file_lock **flp, void **priv) 1327 { 1328 return -EINVAL; 1329 } 1330 1331 static inline int vfs_setlease(struct file *filp, long arg, 1332 struct file_lock **lease, void **priv) 1333 { 1334 return -EINVAL; 1335 } 1336 1337 static inline int lease_modify(struct file_lock *fl, int arg, 1338 struct list_head *dispose) 1339 { 1340 return -EINVAL; 1341 } 1342 1343 struct files_struct; 1344 static inline void show_fd_locks(struct seq_file *f, 1345 struct file *filp, struct files_struct *files) {} 1346 #endif /* !CONFIG_FILE_LOCKING */ 1347 1348 static inline struct inode *file_inode(const struct file *f) 1349 { 1350 return f->f_inode; 1351 } 1352 1353 static inline struct dentry *file_dentry(const struct file *file) 1354 { 1355 return d_real(file->f_path.dentry, file_inode(file)); 1356 } 1357 1358 static inline int locks_lock_file_wait(struct file *filp, struct file_lock *fl) 1359 { 1360 return locks_lock_inode_wait(locks_inode(filp), fl); 1361 } 1362 1363 struct fasync_struct { 1364 rwlock_t fa_lock; 1365 int magic; 1366 int fa_fd; 1367 struct fasync_struct *fa_next; /* singly linked list */ 1368 struct file *fa_file; 1369 struct rcu_head fa_rcu; 1370 }; 1371 1372 #define FASYNC_MAGIC 0x4601 1373 1374 /* SMP safe fasync helpers: */ 1375 extern int fasync_helper(int, struct file *, int, struct fasync_struct **); 1376 extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *); 1377 extern int fasync_remove_entry(struct file *, struct fasync_struct **); 1378 extern struct fasync_struct *fasync_alloc(void); 1379 extern void fasync_free(struct fasync_struct *); 1380 1381 /* can be called from interrupts */ 1382 extern void kill_fasync(struct fasync_struct **, int, int); 1383 1384 extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force); 1385 extern int f_setown(struct file *filp, unsigned long arg, int force); 1386 extern void f_delown(struct file *filp); 1387 extern pid_t f_getown(struct file *filp); 1388 extern int send_sigurg(struct fown_struct *fown); 1389 1390 /* 1391 * sb->s_flags. Note that these mirror the equivalent MS_* flags where 1392 * represented in both. 1393 */ 1394 #define SB_RDONLY 1 /* Mount read-only */ 1395 #define SB_NOSUID 2 /* Ignore suid and sgid bits */ 1396 #define SB_NODEV 4 /* Disallow access to device special files */ 1397 #define SB_NOEXEC 8 /* Disallow program execution */ 1398 #define SB_SYNCHRONOUS 16 /* Writes are synced at once */ 1399 #define SB_MANDLOCK 64 /* Allow mandatory locks on an FS */ 1400 #define SB_DIRSYNC 128 /* Directory modifications are synchronous */ 1401 #define SB_NOATIME 1024 /* Do not update access times. */ 1402 #define SB_NODIRATIME 2048 /* Do not update directory access times */ 1403 #define SB_SILENT 32768 1404 #define SB_POSIXACL (1<<16) /* VFS does not apply the umask */ 1405 #define SB_INLINECRYPT (1<<17) /* Use blk-crypto for encrypted files */ 1406 #define SB_KERNMOUNT (1<<22) /* this is a kern_mount call */ 1407 #define SB_I_VERSION (1<<23) /* Update inode I_version field */ 1408 #define SB_LAZYTIME (1<<25) /* Update the on-disk [acm]times lazily */ 1409 1410 /* These sb flags are internal to the kernel */ 1411 #define SB_SUBMOUNT (1<<26) 1412 #define SB_FORCE (1<<27) 1413 #define SB_NOSEC (1<<28) 1414 #define SB_BORN (1<<29) 1415 #define SB_ACTIVE (1<<30) 1416 #define SB_NOUSER (1<<31) 1417 1418 /* These flags relate to encoding and casefolding */ 1419 #define SB_ENC_STRICT_MODE_FL (1 << 0) 1420 1421 #define sb_has_strict_encoding(sb) \ 1422 (sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL) 1423 1424 /* 1425 * Umount options 1426 */ 1427 1428 #define MNT_FORCE 0x00000001 /* Attempt to forcibily umount */ 1429 #define MNT_DETACH 0x00000002 /* Just detach from the tree */ 1430 #define MNT_EXPIRE 0x00000004 /* Mark for expiry */ 1431 #define UMOUNT_NOFOLLOW 0x00000008 /* Don't follow symlink on umount */ 1432 #define UMOUNT_UNUSED 0x80000000 /* Flag guaranteed to be unused */ 1433 1434 /* sb->s_iflags */ 1435 #define SB_I_CGROUPWB 0x00000001 /* cgroup-aware writeback enabled */ 1436 #define SB_I_NOEXEC 0x00000002 /* Ignore executables on this fs */ 1437 #define SB_I_NODEV 0x00000004 /* Ignore devices on this fs */ 1438 #define SB_I_STABLE_WRITES 0x00000008 /* don't modify blks until WB is done */ 1439 1440 /* sb->s_iflags to limit user namespace mounts */ 1441 #define SB_I_USERNS_VISIBLE 0x00000010 /* fstype already mounted */ 1442 #define SB_I_IMA_UNVERIFIABLE_SIGNATURE 0x00000020 1443 #define SB_I_UNTRUSTED_MOUNTER 0x00000040 1444 1445 #define SB_I_SKIP_SYNC 0x00000100 /* Skip superblock at global sync */ 1446 1447 /* Possible states of 'frozen' field */ 1448 enum { 1449 SB_UNFROZEN = 0, /* FS is unfrozen */ 1450 SB_FREEZE_WRITE = 1, /* Writes, dir ops, ioctls frozen */ 1451 SB_FREEZE_PAGEFAULT = 2, /* Page faults stopped as well */ 1452 SB_FREEZE_FS = 3, /* For internal FS use (e.g. to stop 1453 * internal threads if needed) */ 1454 SB_FREEZE_COMPLETE = 4, /* ->freeze_fs finished successfully */ 1455 }; 1456 1457 #define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1) 1458 1459 struct sb_writers { 1460 int frozen; /* Is sb frozen? */ 1461 wait_queue_head_t wait_unfrozen; /* wait for thaw */ 1462 struct percpu_rw_semaphore rw_sem[SB_FREEZE_LEVELS]; 1463 }; 1464 1465 struct super_block { 1466 struct list_head s_list; /* Keep this first */ 1467 dev_t s_dev; /* search index; _not_ kdev_t */ 1468 unsigned char s_blocksize_bits; 1469 unsigned long s_blocksize; 1470 loff_t s_maxbytes; /* Max file size */ 1471 struct file_system_type *s_type; 1472 const struct super_operations *s_op; 1473 const struct dquot_operations *dq_op; 1474 const struct quotactl_ops *s_qcop; 1475 const struct export_operations *s_export_op; 1476 unsigned long s_flags; 1477 unsigned long s_iflags; /* internal SB_I_* flags */ 1478 unsigned long s_magic; 1479 struct dentry *s_root; 1480 struct rw_semaphore s_umount; 1481 int s_count; 1482 atomic_t s_active; 1483 #ifdef CONFIG_SECURITY 1484 void *s_security; 1485 #endif 1486 const struct xattr_handler **s_xattr; 1487 #ifdef CONFIG_FS_ENCRYPTION 1488 const struct fscrypt_operations *s_cop; 1489 struct key *s_master_keys; /* master crypto keys in use */ 1490 #endif 1491 #ifdef CONFIG_FS_VERITY 1492 const struct fsverity_operations *s_vop; 1493 #endif 1494 #ifdef CONFIG_UNICODE 1495 struct unicode_map *s_encoding; 1496 __u16 s_encoding_flags; 1497 #endif 1498 struct hlist_bl_head s_roots; /* alternate root dentries for NFS */ 1499 struct list_head s_mounts; /* list of mounts; _not_ for fs use */ 1500 struct block_device *s_bdev; 1501 struct backing_dev_info *s_bdi; 1502 struct mtd_info *s_mtd; 1503 struct hlist_node s_instances; 1504 unsigned int s_quota_types; /* Bitmask of supported quota types */ 1505 struct quota_info s_dquot; /* Diskquota specific options */ 1506 1507 struct sb_writers s_writers; 1508 1509 /* 1510 * Keep s_fs_info, s_time_gran, s_fsnotify_mask, and 1511 * s_fsnotify_marks together for cache efficiency. They are frequently 1512 * accessed and rarely modified. 1513 */ 1514 void *s_fs_info; /* Filesystem private info */ 1515 1516 /* Granularity of c/m/atime in ns (cannot be worse than a second) */ 1517 u32 s_time_gran; 1518 /* Time limits for c/m/atime in seconds */ 1519 time64_t s_time_min; 1520 time64_t s_time_max; 1521 #ifdef CONFIG_FSNOTIFY 1522 __u32 s_fsnotify_mask; 1523 struct fsnotify_mark_connector __rcu *s_fsnotify_marks; 1524 #endif 1525 1526 char s_id[32]; /* Informational name */ 1527 uuid_t s_uuid; /* UUID */ 1528 1529 unsigned int s_max_links; 1530 fmode_t s_mode; 1531 1532 /* 1533 * The next field is for VFS *only*. No filesystems have any business 1534 * even looking at it. You had been warned. 1535 */ 1536 struct mutex s_vfs_rename_mutex; /* Kludge */ 1537 1538 /* 1539 * Filesystem subtype. If non-empty the filesystem type field 1540 * in /proc/mounts will be "type.subtype" 1541 */ 1542 const char *s_subtype; 1543 1544 const struct dentry_operations *s_d_op; /* default d_op for dentries */ 1545 1546 /* 1547 * Saved pool identifier for cleancache (-1 means none) 1548 */ 1549 int cleancache_poolid; 1550 1551 struct shrinker s_shrink; /* per-sb shrinker handle */ 1552 1553 /* Number of inodes with nlink == 0 but still referenced */ 1554 atomic_long_t s_remove_count; 1555 1556 /* 1557 * Number of inode/mount/sb objects that are being watched, note that 1558 * inodes objects are currently double-accounted. 1559 */ 1560 atomic_long_t s_fsnotify_connectors; 1561 1562 /* Being remounted read-only */ 1563 int s_readonly_remount; 1564 1565 /* per-sb errseq_t for reporting writeback errors via syncfs */ 1566 errseq_t s_wb_err; 1567 1568 /* AIO completions deferred from interrupt context */ 1569 struct workqueue_struct *s_dio_done_wq; 1570 struct hlist_head s_pins; 1571 1572 /* 1573 * Owning user namespace and default context in which to 1574 * interpret filesystem uids, gids, quotas, device nodes, 1575 * xattrs and security labels. 1576 */ 1577 struct user_namespace *s_user_ns; 1578 1579 /* 1580 * The list_lru structure is essentially just a pointer to a table 1581 * of per-node lru lists, each of which has its own spinlock. 1582 * There is no need to put them into separate cachelines. 1583 */ 1584 struct list_lru s_dentry_lru; 1585 struct list_lru s_inode_lru; 1586 struct rcu_head rcu; 1587 struct work_struct destroy_work; 1588 1589 struct mutex s_sync_lock; /* sync serialisation lock */ 1590 1591 /* 1592 * Indicates how deep in a filesystem stack this SB is 1593 */ 1594 int s_stack_depth; 1595 1596 /* s_inode_list_lock protects s_inodes */ 1597 spinlock_t s_inode_list_lock ____cacheline_aligned_in_smp; 1598 struct list_head s_inodes; /* all inodes */ 1599 1600 spinlock_t s_inode_wblist_lock; 1601 struct list_head s_inodes_wb; /* writeback inodes */ 1602 } __randomize_layout; 1603 1604 /* Helper functions so that in most cases filesystems will 1605 * not need to deal directly with kuid_t and kgid_t and can 1606 * instead deal with the raw numeric values that are stored 1607 * in the filesystem. 1608 */ 1609 static inline uid_t i_uid_read(const struct inode *inode) 1610 { 1611 return from_kuid(inode->i_sb->s_user_ns, inode->i_uid); 1612 } 1613 1614 static inline gid_t i_gid_read(const struct inode *inode) 1615 { 1616 return from_kgid(inode->i_sb->s_user_ns, inode->i_gid); 1617 } 1618 1619 static inline void i_uid_write(struct inode *inode, uid_t uid) 1620 { 1621 inode->i_uid = make_kuid(inode->i_sb->s_user_ns, uid); 1622 } 1623 1624 static inline void i_gid_write(struct inode *inode, gid_t gid) 1625 { 1626 inode->i_gid = make_kgid(inode->i_sb->s_user_ns, gid); 1627 } 1628 1629 /** 1630 * kuid_into_mnt - map a kuid down into a mnt_userns 1631 * @mnt_userns: user namespace of the relevant mount 1632 * @kuid: kuid to be mapped 1633 * 1634 * Return: @kuid mapped according to @mnt_userns. 1635 * If @kuid has no mapping INVALID_UID is returned. 1636 */ 1637 static inline kuid_t kuid_into_mnt(struct user_namespace *mnt_userns, 1638 kuid_t kuid) 1639 { 1640 return make_kuid(mnt_userns, __kuid_val(kuid)); 1641 } 1642 1643 /** 1644 * kgid_into_mnt - map a kgid down into a mnt_userns 1645 * @mnt_userns: user namespace of the relevant mount 1646 * @kgid: kgid to be mapped 1647 * 1648 * Return: @kgid mapped according to @mnt_userns. 1649 * If @kgid has no mapping INVALID_GID is returned. 1650 */ 1651 static inline kgid_t kgid_into_mnt(struct user_namespace *mnt_userns, 1652 kgid_t kgid) 1653 { 1654 return make_kgid(mnt_userns, __kgid_val(kgid)); 1655 } 1656 1657 /** 1658 * i_uid_into_mnt - map an inode's i_uid down into a mnt_userns 1659 * @mnt_userns: user namespace of the mount the inode was found from 1660 * @inode: inode to map 1661 * 1662 * Return: the inode's i_uid mapped down according to @mnt_userns. 1663 * If the inode's i_uid has no mapping INVALID_UID is returned. 1664 */ 1665 static inline kuid_t i_uid_into_mnt(struct user_namespace *mnt_userns, 1666 const struct inode *inode) 1667 { 1668 return kuid_into_mnt(mnt_userns, inode->i_uid); 1669 } 1670 1671 /** 1672 * i_gid_into_mnt - map an inode's i_gid down into a mnt_userns 1673 * @mnt_userns: user namespace of the mount the inode was found from 1674 * @inode: inode to map 1675 * 1676 * Return: the inode's i_gid mapped down according to @mnt_userns. 1677 * If the inode's i_gid has no mapping INVALID_GID is returned. 1678 */ 1679 static inline kgid_t i_gid_into_mnt(struct user_namespace *mnt_userns, 1680 const struct inode *inode) 1681 { 1682 return kgid_into_mnt(mnt_userns, inode->i_gid); 1683 } 1684 1685 /** 1686 * kuid_from_mnt - map a kuid up into a mnt_userns 1687 * @mnt_userns: user namespace of the relevant mount 1688 * @kuid: kuid to be mapped 1689 * 1690 * Return: @kuid mapped up according to @mnt_userns. 1691 * If @kuid has no mapping INVALID_UID is returned. 1692 */ 1693 static inline kuid_t kuid_from_mnt(struct user_namespace *mnt_userns, 1694 kuid_t kuid) 1695 { 1696 return KUIDT_INIT(from_kuid(mnt_userns, kuid)); 1697 } 1698 1699 /** 1700 * kgid_from_mnt - map a kgid up into a mnt_userns 1701 * @mnt_userns: user namespace of the relevant mount 1702 * @kgid: kgid to be mapped 1703 * 1704 * Return: @kgid mapped up according to @mnt_userns. 1705 * If @kgid has no mapping INVALID_GID is returned. 1706 */ 1707 static inline kgid_t kgid_from_mnt(struct user_namespace *mnt_userns, 1708 kgid_t kgid) 1709 { 1710 return KGIDT_INIT(from_kgid(mnt_userns, kgid)); 1711 } 1712 1713 /** 1714 * mapped_fsuid - return caller's fsuid mapped up into a mnt_userns 1715 * @mnt_userns: user namespace of the relevant mount 1716 * 1717 * Use this helper to initialize a new vfs or filesystem object based on 1718 * the caller's fsuid. A common example is initializing the i_uid field of 1719 * a newly allocated inode triggered by a creation event such as mkdir or 1720 * O_CREAT. Other examples include the allocation of quotas for a specific 1721 * user. 1722 * 1723 * Return: the caller's current fsuid mapped up according to @mnt_userns. 1724 */ 1725 static inline kuid_t mapped_fsuid(struct user_namespace *mnt_userns) 1726 { 1727 return kuid_from_mnt(mnt_userns, current_fsuid()); 1728 } 1729 1730 /** 1731 * mapped_fsgid - return caller's fsgid mapped up into a mnt_userns 1732 * @mnt_userns: user namespace of the relevant mount 1733 * 1734 * Use this helper to initialize a new vfs or filesystem object based on 1735 * the caller's fsgid. A common example is initializing the i_gid field of 1736 * a newly allocated inode triggered by a creation event such as mkdir or 1737 * O_CREAT. Other examples include the allocation of quotas for a specific 1738 * user. 1739 * 1740 * Return: the caller's current fsgid mapped up according to @mnt_userns. 1741 */ 1742 static inline kgid_t mapped_fsgid(struct user_namespace *mnt_userns) 1743 { 1744 return kgid_from_mnt(mnt_userns, current_fsgid()); 1745 } 1746 1747 /** 1748 * inode_fsuid_set - initialize inode's i_uid field with callers fsuid 1749 * @inode: inode to initialize 1750 * @mnt_userns: user namespace of the mount the inode was found from 1751 * 1752 * Initialize the i_uid field of @inode. If the inode was found/created via 1753 * an idmapped mount map the caller's fsuid according to @mnt_users. 1754 */ 1755 static inline void inode_fsuid_set(struct inode *inode, 1756 struct user_namespace *mnt_userns) 1757 { 1758 inode->i_uid = mapped_fsuid(mnt_userns); 1759 } 1760 1761 /** 1762 * inode_fsgid_set - initialize inode's i_gid field with callers fsgid 1763 * @inode: inode to initialize 1764 * @mnt_userns: user namespace of the mount the inode was found from 1765 * 1766 * Initialize the i_gid field of @inode. If the inode was found/created via 1767 * an idmapped mount map the caller's fsgid according to @mnt_users. 1768 */ 1769 static inline void inode_fsgid_set(struct inode *inode, 1770 struct user_namespace *mnt_userns) 1771 { 1772 inode->i_gid = mapped_fsgid(mnt_userns); 1773 } 1774 1775 /** 1776 * fsuidgid_has_mapping() - check whether caller's fsuid/fsgid is mapped 1777 * @sb: the superblock we want a mapping in 1778 * @mnt_userns: user namespace of the relevant mount 1779 * 1780 * Check whether the caller's fsuid and fsgid have a valid mapping in the 1781 * s_user_ns of the superblock @sb. If the caller is on an idmapped mount map 1782 * the caller's fsuid and fsgid according to the @mnt_userns first. 1783 * 1784 * Return: true if fsuid and fsgid is mapped, false if not. 1785 */ 1786 static inline bool fsuidgid_has_mapping(struct super_block *sb, 1787 struct user_namespace *mnt_userns) 1788 { 1789 struct user_namespace *s_user_ns = sb->s_user_ns; 1790 1791 return kuid_has_mapping(s_user_ns, mapped_fsuid(mnt_userns)) && 1792 kgid_has_mapping(s_user_ns, mapped_fsgid(mnt_userns)); 1793 } 1794 1795 extern struct timespec64 current_time(struct inode *inode); 1796 1797 /* 1798 * Snapshotting support. 1799 */ 1800 1801 /* 1802 * These are internal functions, please use sb_start_{write,pagefault,intwrite} 1803 * instead. 1804 */ 1805 static inline void __sb_end_write(struct super_block *sb, int level) 1806 { 1807 percpu_up_read(sb->s_writers.rw_sem + level-1); 1808 } 1809 1810 static inline void __sb_start_write(struct super_block *sb, int level) 1811 { 1812 percpu_down_read(sb->s_writers.rw_sem + level - 1); 1813 } 1814 1815 static inline bool __sb_start_write_trylock(struct super_block *sb, int level) 1816 { 1817 return percpu_down_read_trylock(sb->s_writers.rw_sem + level - 1); 1818 } 1819 1820 #define __sb_writers_acquired(sb, lev) \ 1821 percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_) 1822 #define __sb_writers_release(sb, lev) \ 1823 percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_) 1824 1825 /** 1826 * sb_end_write - drop write access to a superblock 1827 * @sb: the super we wrote to 1828 * 1829 * Decrement number of writers to the filesystem. Wake up possible waiters 1830 * wanting to freeze the filesystem. 1831 */ 1832 static inline void sb_end_write(struct super_block *sb) 1833 { 1834 __sb_end_write(sb, SB_FREEZE_WRITE); 1835 } 1836 1837 /** 1838 * sb_end_pagefault - drop write access to a superblock from a page fault 1839 * @sb: the super we wrote to 1840 * 1841 * Decrement number of processes handling write page fault to the filesystem. 1842 * Wake up possible waiters wanting to freeze the filesystem. 1843 */ 1844 static inline void sb_end_pagefault(struct super_block *sb) 1845 { 1846 __sb_end_write(sb, SB_FREEZE_PAGEFAULT); 1847 } 1848 1849 /** 1850 * sb_end_intwrite - drop write access to a superblock for internal fs purposes 1851 * @sb: the super we wrote to 1852 * 1853 * Decrement fs-internal number of writers to the filesystem. Wake up possible 1854 * waiters wanting to freeze the filesystem. 1855 */ 1856 static inline void sb_end_intwrite(struct super_block *sb) 1857 { 1858 __sb_end_write(sb, SB_FREEZE_FS); 1859 } 1860 1861 /** 1862 * sb_start_write - get write access to a superblock 1863 * @sb: the super we write to 1864 * 1865 * When a process wants to write data or metadata to a file system (i.e. dirty 1866 * a page or an inode), it should embed the operation in a sb_start_write() - 1867 * sb_end_write() pair to get exclusion against file system freezing. This 1868 * function increments number of writers preventing freezing. If the file 1869 * system is already frozen, the function waits until the file system is 1870 * thawed. 1871 * 1872 * Since freeze protection behaves as a lock, users have to preserve 1873 * ordering of freeze protection and other filesystem locks. Generally, 1874 * freeze protection should be the outermost lock. In particular, we have: 1875 * 1876 * sb_start_write 1877 * -> i_mutex (write path, truncate, directory ops, ...) 1878 * -> s_umount (freeze_super, thaw_super) 1879 */ 1880 static inline void sb_start_write(struct super_block *sb) 1881 { 1882 __sb_start_write(sb, SB_FREEZE_WRITE); 1883 } 1884 1885 static inline bool sb_start_write_trylock(struct super_block *sb) 1886 { 1887 return __sb_start_write_trylock(sb, SB_FREEZE_WRITE); 1888 } 1889 1890 /** 1891 * sb_start_pagefault - get write access to a superblock from a page fault 1892 * @sb: the super we write to 1893 * 1894 * When a process starts handling write page fault, it should embed the 1895 * operation into sb_start_pagefault() - sb_end_pagefault() pair to get 1896 * exclusion against file system freezing. This is needed since the page fault 1897 * is going to dirty a page. This function increments number of running page 1898 * faults preventing freezing. If the file system is already frozen, the 1899 * function waits until the file system is thawed. 1900 * 1901 * Since page fault freeze protection behaves as a lock, users have to preserve 1902 * ordering of freeze protection and other filesystem locks. It is advised to 1903 * put sb_start_pagefault() close to mmap_lock in lock ordering. Page fault 1904 * handling code implies lock dependency: 1905 * 1906 * mmap_lock 1907 * -> sb_start_pagefault 1908 */ 1909 static inline void sb_start_pagefault(struct super_block *sb) 1910 { 1911 __sb_start_write(sb, SB_FREEZE_PAGEFAULT); 1912 } 1913 1914 /** 1915 * sb_start_intwrite - get write access to a superblock for internal fs purposes 1916 * @sb: the super we write to 1917 * 1918 * This is the third level of protection against filesystem freezing. It is 1919 * free for use by a filesystem. The only requirement is that it must rank 1920 * below sb_start_pagefault. 1921 * 1922 * For example filesystem can call sb_start_intwrite() when starting a 1923 * transaction which somewhat eases handling of freezing for internal sources 1924 * of filesystem changes (internal fs threads, discarding preallocation on file 1925 * close, etc.). 1926 */ 1927 static inline void sb_start_intwrite(struct super_block *sb) 1928 { 1929 __sb_start_write(sb, SB_FREEZE_FS); 1930 } 1931 1932 static inline bool sb_start_intwrite_trylock(struct super_block *sb) 1933 { 1934 return __sb_start_write_trylock(sb, SB_FREEZE_FS); 1935 } 1936 1937 bool inode_owner_or_capable(struct user_namespace *mnt_userns, 1938 const struct inode *inode); 1939 1940 /* 1941 * VFS helper functions.. 1942 */ 1943 int vfs_create(struct user_namespace *, struct inode *, 1944 struct dentry *, umode_t, bool); 1945 int vfs_mkdir(struct user_namespace *, struct inode *, 1946 struct dentry *, umode_t); 1947 int vfs_mknod(struct user_namespace *, struct inode *, struct dentry *, 1948 umode_t, dev_t); 1949 int vfs_symlink(struct user_namespace *, struct inode *, 1950 struct dentry *, const char *); 1951 int vfs_link(struct dentry *, struct user_namespace *, struct inode *, 1952 struct dentry *, struct inode **); 1953 int vfs_rmdir(struct user_namespace *, struct inode *, struct dentry *); 1954 int vfs_unlink(struct user_namespace *, struct inode *, struct dentry *, 1955 struct inode **); 1956 1957 /** 1958 * struct renamedata - contains all information required for renaming 1959 * @old_mnt_userns: old user namespace of the mount the inode was found from 1960 * @old_dir: parent of source 1961 * @old_dentry: source 1962 * @new_mnt_userns: new user namespace of the mount the inode was found from 1963 * @new_dir: parent of destination 1964 * @new_dentry: destination 1965 * @delegated_inode: returns an inode needing a delegation break 1966 * @flags: rename flags 1967 */ 1968 struct renamedata { 1969 struct user_namespace *old_mnt_userns; 1970 struct inode *old_dir; 1971 struct dentry *old_dentry; 1972 struct user_namespace *new_mnt_userns; 1973 struct inode *new_dir; 1974 struct dentry *new_dentry; 1975 struct inode **delegated_inode; 1976 unsigned int flags; 1977 } __randomize_layout; 1978 1979 int vfs_rename(struct renamedata *); 1980 1981 static inline int vfs_whiteout(struct user_namespace *mnt_userns, 1982 struct inode *dir, struct dentry *dentry) 1983 { 1984 return vfs_mknod(mnt_userns, dir, dentry, S_IFCHR | WHITEOUT_MODE, 1985 WHITEOUT_DEV); 1986 } 1987 1988 struct dentry *vfs_tmpfile(struct user_namespace *mnt_userns, 1989 struct dentry *dentry, umode_t mode, int open_flag); 1990 1991 int vfs_mkobj(struct dentry *, umode_t, 1992 int (*f)(struct dentry *, umode_t, void *), 1993 void *); 1994 1995 int vfs_fchown(struct file *file, uid_t user, gid_t group); 1996 int vfs_fchmod(struct file *file, umode_t mode); 1997 int vfs_utimes(const struct path *path, struct timespec64 *times); 1998 1999 extern long vfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 2000 2001 #ifdef CONFIG_COMPAT 2002 extern long compat_ptr_ioctl(struct file *file, unsigned int cmd, 2003 unsigned long arg); 2004 #else 2005 #define compat_ptr_ioctl NULL 2006 #endif 2007 2008 /* 2009 * VFS file helper functions. 2010 */ 2011 void inode_init_owner(struct user_namespace *mnt_userns, struct inode *inode, 2012 const struct inode *dir, umode_t mode); 2013 extern bool may_open_dev(const struct path *path); 2014 2015 /* 2016 * This is the "filldir" function type, used by readdir() to let 2017 * the kernel specify what kind of dirent layout it wants to have. 2018 * This allows the kernel to read directories into kernel space or 2019 * to have different dirent layouts depending on the binary type. 2020 */ 2021 struct dir_context; 2022 typedef int (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64, 2023 unsigned); 2024 2025 struct dir_context { 2026 filldir_t actor; 2027 loff_t pos; 2028 }; 2029 2030 /* 2031 * These flags let !MMU mmap() govern direct device mapping vs immediate 2032 * copying more easily for MAP_PRIVATE, especially for ROM filesystems. 2033 * 2034 * NOMMU_MAP_COPY: Copy can be mapped (MAP_PRIVATE) 2035 * NOMMU_MAP_DIRECT: Can be mapped directly (MAP_SHARED) 2036 * NOMMU_MAP_READ: Can be mapped for reading 2037 * NOMMU_MAP_WRITE: Can be mapped for writing 2038 * NOMMU_MAP_EXEC: Can be mapped for execution 2039 */ 2040 #define NOMMU_MAP_COPY 0x00000001 2041 #define NOMMU_MAP_DIRECT 0x00000008 2042 #define NOMMU_MAP_READ VM_MAYREAD 2043 #define NOMMU_MAP_WRITE VM_MAYWRITE 2044 #define NOMMU_MAP_EXEC VM_MAYEXEC 2045 2046 #define NOMMU_VMFLAGS \ 2047 (NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC) 2048 2049 /* 2050 * These flags control the behavior of the remap_file_range function pointer. 2051 * If it is called with len == 0 that means "remap to end of source file". 2052 * See Documentation/filesystems/vfs.rst for more details about this call. 2053 * 2054 * REMAP_FILE_DEDUP: only remap if contents identical (i.e. deduplicate) 2055 * REMAP_FILE_CAN_SHORTEN: caller can handle a shortened request 2056 */ 2057 #define REMAP_FILE_DEDUP (1 << 0) 2058 #define REMAP_FILE_CAN_SHORTEN (1 << 1) 2059 2060 /* 2061 * These flags signal that the caller is ok with altering various aspects of 2062 * the behavior of the remap operation. The changes must be made by the 2063 * implementation; the vfs remap helper functions can take advantage of them. 2064 * Flags in this category exist to preserve the quirky behavior of the hoisted 2065 * btrfs clone/dedupe ioctls. 2066 */ 2067 #define REMAP_FILE_ADVISORY (REMAP_FILE_CAN_SHORTEN) 2068 2069 struct iov_iter; 2070 2071 struct file_operations { 2072 struct module *owner; 2073 loff_t (*llseek) (struct file *, loff_t, int); 2074 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *); 2075 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *); 2076 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *); 2077 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *); 2078 int (*iopoll)(struct kiocb *kiocb, bool spin); 2079 int (*iterate) (struct file *, struct dir_context *); 2080 int (*iterate_shared) (struct file *, struct dir_context *); 2081 __poll_t (*poll) (struct file *, struct poll_table_struct *); 2082 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long); 2083 long (*compat_ioctl) (struct file *, unsigned int, unsigned long); 2084 int (*mmap) (struct file *, struct vm_area_struct *); 2085 unsigned long mmap_supported_flags; 2086 int (*open) (struct inode *, struct file *); 2087 int (*flush) (struct file *, fl_owner_t id); 2088 int (*release) (struct inode *, struct file *); 2089 int (*fsync) (struct file *, loff_t, loff_t, int datasync); 2090 int (*fasync) (int, struct file *, int); 2091 int (*lock) (struct file *, int, struct file_lock *); 2092 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int); 2093 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 2094 int (*check_flags)(int); 2095 int (*flock) (struct file *, int, struct file_lock *); 2096 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int); 2097 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int); 2098 int (*setlease)(struct file *, long, struct file_lock **, void **); 2099 long (*fallocate)(struct file *file, int mode, loff_t offset, 2100 loff_t len); 2101 void (*show_fdinfo)(struct seq_file *m, struct file *f); 2102 #ifndef CONFIG_MMU 2103 unsigned (*mmap_capabilities)(struct file *); 2104 #endif 2105 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *, 2106 loff_t, size_t, unsigned int); 2107 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in, 2108 struct file *file_out, loff_t pos_out, 2109 loff_t len, unsigned int remap_flags); 2110 int (*fadvise)(struct file *, loff_t, loff_t, int); 2111 } __randomize_layout; 2112 2113 struct inode_operations { 2114 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int); 2115 const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *); 2116 int (*permission) (struct user_namespace *, struct inode *, int); 2117 struct posix_acl * (*get_acl)(struct inode *, int, bool); 2118 2119 int (*readlink) (struct dentry *, char __user *,int); 2120 2121 int (*create) (struct user_namespace *, struct inode *,struct dentry *, 2122 umode_t, bool); 2123 int (*link) (struct dentry *,struct inode *,struct dentry *); 2124 int (*unlink) (struct inode *,struct dentry *); 2125 int (*symlink) (struct user_namespace *, struct inode *,struct dentry *, 2126 const char *); 2127 int (*mkdir) (struct user_namespace *, struct inode *,struct dentry *, 2128 umode_t); 2129 int (*rmdir) (struct inode *,struct dentry *); 2130 int (*mknod) (struct user_namespace *, struct inode *,struct dentry *, 2131 umode_t,dev_t); 2132 int (*rename) (struct user_namespace *, struct inode *, struct dentry *, 2133 struct inode *, struct dentry *, unsigned int); 2134 int (*setattr) (struct user_namespace *, struct dentry *, 2135 struct iattr *); 2136 int (*getattr) (struct user_namespace *, const struct path *, 2137 struct kstat *, u32, unsigned int); 2138 ssize_t (*listxattr) (struct dentry *, char *, size_t); 2139 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start, 2140 u64 len); 2141 int (*update_time)(struct inode *, struct timespec64 *, int); 2142 int (*atomic_open)(struct inode *, struct dentry *, 2143 struct file *, unsigned open_flag, 2144 umode_t create_mode); 2145 int (*tmpfile) (struct user_namespace *, struct inode *, 2146 struct dentry *, umode_t); 2147 int (*set_acl)(struct user_namespace *, struct inode *, 2148 struct posix_acl *, int); 2149 int (*fileattr_set)(struct user_namespace *mnt_userns, 2150 struct dentry *dentry, struct fileattr *fa); 2151 int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa); 2152 } ____cacheline_aligned; 2153 2154 static inline ssize_t call_read_iter(struct file *file, struct kiocb *kio, 2155 struct iov_iter *iter) 2156 { 2157 return file->f_op->read_iter(kio, iter); 2158 } 2159 2160 static inline ssize_t call_write_iter(struct file *file, struct kiocb *kio, 2161 struct iov_iter *iter) 2162 { 2163 return file->f_op->write_iter(kio, iter); 2164 } 2165 2166 static inline int call_mmap(struct file *file, struct vm_area_struct *vma) 2167 { 2168 return file->f_op->mmap(file, vma); 2169 } 2170 2171 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *); 2172 extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *); 2173 extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *, 2174 loff_t, size_t, unsigned int); 2175 extern ssize_t generic_copy_file_range(struct file *file_in, loff_t pos_in, 2176 struct file *file_out, loff_t pos_out, 2177 size_t len, unsigned int flags); 2178 extern int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in, 2179 struct file *file_out, loff_t pos_out, 2180 loff_t *count, 2181 unsigned int remap_flags); 2182 extern loff_t do_clone_file_range(struct file *file_in, loff_t pos_in, 2183 struct file *file_out, loff_t pos_out, 2184 loff_t len, unsigned int remap_flags); 2185 extern loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in, 2186 struct file *file_out, loff_t pos_out, 2187 loff_t len, unsigned int remap_flags); 2188 extern int vfs_dedupe_file_range(struct file *file, 2189 struct file_dedupe_range *same); 2190 extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos, 2191 struct file *dst_file, loff_t dst_pos, 2192 loff_t len, unsigned int remap_flags); 2193 2194 2195 struct super_operations { 2196 struct inode *(*alloc_inode)(struct super_block *sb); 2197 void (*destroy_inode)(struct inode *); 2198 void (*free_inode)(struct inode *); 2199 2200 void (*dirty_inode) (struct inode *, int flags); 2201 int (*write_inode) (struct inode *, struct writeback_control *wbc); 2202 int (*drop_inode) (struct inode *); 2203 void (*evict_inode) (struct inode *); 2204 void (*put_super) (struct super_block *); 2205 int (*sync_fs)(struct super_block *sb, int wait); 2206 int (*freeze_super) (struct super_block *); 2207 int (*freeze_fs) (struct super_block *); 2208 int (*thaw_super) (struct super_block *); 2209 int (*unfreeze_fs) (struct super_block *); 2210 int (*statfs) (struct dentry *, struct kstatfs *); 2211 int (*remount_fs) (struct super_block *, int *, char *); 2212 void (*umount_begin) (struct super_block *); 2213 2214 int (*show_options)(struct seq_file *, struct dentry *); 2215 int (*show_devname)(struct seq_file *, struct dentry *); 2216 int (*show_path)(struct seq_file *, struct dentry *); 2217 int (*show_stats)(struct seq_file *, struct dentry *); 2218 #ifdef CONFIG_QUOTA 2219 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t); 2220 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t); 2221 struct dquot **(*get_dquots)(struct inode *); 2222 #endif 2223 long (*nr_cached_objects)(struct super_block *, 2224 struct shrink_control *); 2225 long (*free_cached_objects)(struct super_block *, 2226 struct shrink_control *); 2227 }; 2228 2229 /* 2230 * Inode flags - they have no relation to superblock flags now 2231 */ 2232 #define S_SYNC (1 << 0) /* Writes are synced at once */ 2233 #define S_NOATIME (1 << 1) /* Do not update access times */ 2234 #define S_APPEND (1 << 2) /* Append-only file */ 2235 #define S_IMMUTABLE (1 << 3) /* Immutable file */ 2236 #define S_DEAD (1 << 4) /* removed, but still open directory */ 2237 #define S_NOQUOTA (1 << 5) /* Inode is not counted to quota */ 2238 #define S_DIRSYNC (1 << 6) /* Directory modifications are synchronous */ 2239 #define S_NOCMTIME (1 << 7) /* Do not update file c/mtime */ 2240 #define S_SWAPFILE (1 << 8) /* Do not truncate: swapon got its bmaps */ 2241 #define S_PRIVATE (1 << 9) /* Inode is fs-internal */ 2242 #define S_IMA (1 << 10) /* Inode has an associated IMA struct */ 2243 #define S_AUTOMOUNT (1 << 11) /* Automount/referral quasi-directory */ 2244 #define S_NOSEC (1 << 12) /* no suid or xattr security attributes */ 2245 #ifdef CONFIG_FS_DAX 2246 #define S_DAX (1 << 13) /* Direct Access, avoiding the page cache */ 2247 #else 2248 #define S_DAX 0 /* Make all the DAX code disappear */ 2249 #endif 2250 #define S_ENCRYPTED (1 << 14) /* Encrypted file (using fs/crypto/) */ 2251 #define S_CASEFOLD (1 << 15) /* Casefolded file */ 2252 #define S_VERITY (1 << 16) /* Verity file (using fs/verity/) */ 2253 2254 /* 2255 * Note that nosuid etc flags are inode-specific: setting some file-system 2256 * flags just means all the inodes inherit those flags by default. It might be 2257 * possible to override it selectively if you really wanted to with some 2258 * ioctl() that is not currently implemented. 2259 * 2260 * Exception: SB_RDONLY is always applied to the entire file system. 2261 * 2262 * Unfortunately, it is possible to change a filesystems flags with it mounted 2263 * with files in use. This means that all of the inodes will not have their 2264 * i_flags updated. Hence, i_flags no longer inherit the superblock mount 2265 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org 2266 */ 2267 #define __IS_FLG(inode, flg) ((inode)->i_sb->s_flags & (flg)) 2268 2269 static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; } 2270 #define IS_RDONLY(inode) sb_rdonly((inode)->i_sb) 2271 #define IS_SYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS) || \ 2272 ((inode)->i_flags & S_SYNC)) 2273 #define IS_DIRSYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS|SB_DIRSYNC) || \ 2274 ((inode)->i_flags & (S_SYNC|S_DIRSYNC))) 2275 #define IS_MANDLOCK(inode) __IS_FLG(inode, SB_MANDLOCK) 2276 #define IS_NOATIME(inode) __IS_FLG(inode, SB_RDONLY|SB_NOATIME) 2277 #define IS_I_VERSION(inode) __IS_FLG(inode, SB_I_VERSION) 2278 2279 #define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA) 2280 #define IS_APPEND(inode) ((inode)->i_flags & S_APPEND) 2281 #define IS_IMMUTABLE(inode) ((inode)->i_flags & S_IMMUTABLE) 2282 #define IS_POSIXACL(inode) __IS_FLG(inode, SB_POSIXACL) 2283 2284 #define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD) 2285 #define IS_NOCMTIME(inode) ((inode)->i_flags & S_NOCMTIME) 2286 #define IS_SWAPFILE(inode) ((inode)->i_flags & S_SWAPFILE) 2287 #define IS_PRIVATE(inode) ((inode)->i_flags & S_PRIVATE) 2288 #define IS_IMA(inode) ((inode)->i_flags & S_IMA) 2289 #define IS_AUTOMOUNT(inode) ((inode)->i_flags & S_AUTOMOUNT) 2290 #define IS_NOSEC(inode) ((inode)->i_flags & S_NOSEC) 2291 #define IS_DAX(inode) ((inode)->i_flags & S_DAX) 2292 #define IS_ENCRYPTED(inode) ((inode)->i_flags & S_ENCRYPTED) 2293 #define IS_CASEFOLDED(inode) ((inode)->i_flags & S_CASEFOLD) 2294 #define IS_VERITY(inode) ((inode)->i_flags & S_VERITY) 2295 2296 #define IS_WHITEOUT(inode) (S_ISCHR(inode->i_mode) && \ 2297 (inode)->i_rdev == WHITEOUT_DEV) 2298 2299 static inline bool HAS_UNMAPPED_ID(struct user_namespace *mnt_userns, 2300 struct inode *inode) 2301 { 2302 return !uid_valid(i_uid_into_mnt(mnt_userns, inode)) || 2303 !gid_valid(i_gid_into_mnt(mnt_userns, inode)); 2304 } 2305 2306 static inline enum rw_hint file_write_hint(struct file *file) 2307 { 2308 if (file->f_write_hint != WRITE_LIFE_NOT_SET) 2309 return file->f_write_hint; 2310 2311 return file_inode(file)->i_write_hint; 2312 } 2313 2314 static inline int iocb_flags(struct file *file); 2315 2316 static inline u16 ki_hint_validate(enum rw_hint hint) 2317 { 2318 typeof(((struct kiocb *)0)->ki_hint) max_hint = -1; 2319 2320 if (hint <= max_hint) 2321 return hint; 2322 return 0; 2323 } 2324 2325 static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp) 2326 { 2327 *kiocb = (struct kiocb) { 2328 .ki_filp = filp, 2329 .ki_flags = iocb_flags(filp), 2330 .ki_hint = ki_hint_validate(file_write_hint(filp)), 2331 .ki_ioprio = get_current_ioprio(), 2332 }; 2333 } 2334 2335 static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src, 2336 struct file *filp) 2337 { 2338 *kiocb = (struct kiocb) { 2339 .ki_filp = filp, 2340 .ki_flags = kiocb_src->ki_flags, 2341 .ki_hint = kiocb_src->ki_hint, 2342 .ki_ioprio = kiocb_src->ki_ioprio, 2343 .ki_pos = kiocb_src->ki_pos, 2344 }; 2345 } 2346 2347 /* 2348 * Inode state bits. Protected by inode->i_lock 2349 * 2350 * Four bits determine the dirty state of the inode: I_DIRTY_SYNC, 2351 * I_DIRTY_DATASYNC, I_DIRTY_PAGES, and I_DIRTY_TIME. 2352 * 2353 * Four bits define the lifetime of an inode. Initially, inodes are I_NEW, 2354 * until that flag is cleared. I_WILL_FREE, I_FREEING and I_CLEAR are set at 2355 * various stages of removing an inode. 2356 * 2357 * Two bits are used for locking and completion notification, I_NEW and I_SYNC. 2358 * 2359 * I_DIRTY_SYNC Inode is dirty, but doesn't have to be written on 2360 * fdatasync() (unless I_DIRTY_DATASYNC is also set). 2361 * Timestamp updates are the usual cause. 2362 * I_DIRTY_DATASYNC Data-related inode changes pending. We keep track of 2363 * these changes separately from I_DIRTY_SYNC so that we 2364 * don't have to write inode on fdatasync() when only 2365 * e.g. the timestamps have changed. 2366 * I_DIRTY_PAGES Inode has dirty pages. Inode itself may be clean. 2367 * I_DIRTY_TIME The inode itself only has dirty timestamps, and the 2368 * lazytime mount option is enabled. We keep track of this 2369 * separately from I_DIRTY_SYNC in order to implement 2370 * lazytime. This gets cleared if I_DIRTY_INODE 2371 * (I_DIRTY_SYNC and/or I_DIRTY_DATASYNC) gets set. I.e. 2372 * either I_DIRTY_TIME *or* I_DIRTY_INODE can be set in 2373 * i_state, but not both. I_DIRTY_PAGES may still be set. 2374 * I_NEW Serves as both a mutex and completion notification. 2375 * New inodes set I_NEW. If two processes both create 2376 * the same inode, one of them will release its inode and 2377 * wait for I_NEW to be released before returning. 2378 * Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can 2379 * also cause waiting on I_NEW, without I_NEW actually 2380 * being set. find_inode() uses this to prevent returning 2381 * nearly-dead inodes. 2382 * I_WILL_FREE Must be set when calling write_inode_now() if i_count 2383 * is zero. I_FREEING must be set when I_WILL_FREE is 2384 * cleared. 2385 * I_FREEING Set when inode is about to be freed but still has dirty 2386 * pages or buffers attached or the inode itself is still 2387 * dirty. 2388 * I_CLEAR Added by clear_inode(). In this state the inode is 2389 * clean and can be destroyed. Inode keeps I_FREEING. 2390 * 2391 * Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are 2392 * prohibited for many purposes. iget() must wait for 2393 * the inode to be completely released, then create it 2394 * anew. Other functions will just ignore such inodes, 2395 * if appropriate. I_NEW is used for waiting. 2396 * 2397 * I_SYNC Writeback of inode is running. The bit is set during 2398 * data writeback, and cleared with a wakeup on the bit 2399 * address once it is done. The bit is also used to pin 2400 * the inode in memory for flusher thread. 2401 * 2402 * I_REFERENCED Marks the inode as recently references on the LRU list. 2403 * 2404 * I_DIO_WAKEUP Never set. Only used as a key for wait_on_bit(). 2405 * 2406 * I_WB_SWITCH Cgroup bdi_writeback switching in progress. Used to 2407 * synchronize competing switching instances and to tell 2408 * wb stat updates to grab the i_pages lock. See 2409 * inode_switch_wbs_work_fn() for details. 2410 * 2411 * I_OVL_INUSE Used by overlayfs to get exclusive ownership on upper 2412 * and work dirs among overlayfs mounts. 2413 * 2414 * I_CREATING New object's inode in the middle of setting up. 2415 * 2416 * I_DONTCACHE Evict inode as soon as it is not used anymore. 2417 * 2418 * I_SYNC_QUEUED Inode is queued in b_io or b_more_io writeback lists. 2419 * Used to detect that mark_inode_dirty() should not move 2420 * inode between dirty lists. 2421 * 2422 * Q: What is the difference between I_WILL_FREE and I_FREEING? 2423 */ 2424 #define I_DIRTY_SYNC (1 << 0) 2425 #define I_DIRTY_DATASYNC (1 << 1) 2426 #define I_DIRTY_PAGES (1 << 2) 2427 #define __I_NEW 3 2428 #define I_NEW (1 << __I_NEW) 2429 #define I_WILL_FREE (1 << 4) 2430 #define I_FREEING (1 << 5) 2431 #define I_CLEAR (1 << 6) 2432 #define __I_SYNC 7 2433 #define I_SYNC (1 << __I_SYNC) 2434 #define I_REFERENCED (1 << 8) 2435 #define __I_DIO_WAKEUP 9 2436 #define I_DIO_WAKEUP (1 << __I_DIO_WAKEUP) 2437 #define I_LINKABLE (1 << 10) 2438 #define I_DIRTY_TIME (1 << 11) 2439 #define I_WB_SWITCH (1 << 13) 2440 #define I_OVL_INUSE (1 << 14) 2441 #define I_CREATING (1 << 15) 2442 #define I_DONTCACHE (1 << 16) 2443 #define I_SYNC_QUEUED (1 << 17) 2444 2445 #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC) 2446 #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES) 2447 #define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME) 2448 2449 extern void __mark_inode_dirty(struct inode *, int); 2450 static inline void mark_inode_dirty(struct inode *inode) 2451 { 2452 __mark_inode_dirty(inode, I_DIRTY); 2453 } 2454 2455 static inline void mark_inode_dirty_sync(struct inode *inode) 2456 { 2457 __mark_inode_dirty(inode, I_DIRTY_SYNC); 2458 } 2459 2460 /* 2461 * Returns true if the given inode itself only has dirty timestamps (its pages 2462 * may still be dirty) and isn't currently being allocated or freed. 2463 * Filesystems should call this if when writing an inode when lazytime is 2464 * enabled, they want to opportunistically write the timestamps of other inodes 2465 * located very nearby on-disk, e.g. in the same inode block. This returns true 2466 * if the given inode is in need of such an opportunistic update. Requires 2467 * i_lock, or at least later re-checking under i_lock. 2468 */ 2469 static inline bool inode_is_dirtytime_only(struct inode *inode) 2470 { 2471 return (inode->i_state & (I_DIRTY_TIME | I_NEW | 2472 I_FREEING | I_WILL_FREE)) == I_DIRTY_TIME; 2473 } 2474 2475 extern void inc_nlink(struct inode *inode); 2476 extern void drop_nlink(struct inode *inode); 2477 extern void clear_nlink(struct inode *inode); 2478 extern void set_nlink(struct inode *inode, unsigned int nlink); 2479 2480 static inline void inode_inc_link_count(struct inode *inode) 2481 { 2482 inc_nlink(inode); 2483 mark_inode_dirty(inode); 2484 } 2485 2486 static inline void inode_dec_link_count(struct inode *inode) 2487 { 2488 drop_nlink(inode); 2489 mark_inode_dirty(inode); 2490 } 2491 2492 enum file_time_flags { 2493 S_ATIME = 1, 2494 S_MTIME = 2, 2495 S_CTIME = 4, 2496 S_VERSION = 8, 2497 }; 2498 2499 extern bool atime_needs_update(const struct path *, struct inode *); 2500 extern void touch_atime(const struct path *); 2501 static inline void file_accessed(struct file *file) 2502 { 2503 if (!(file->f_flags & O_NOATIME)) 2504 touch_atime(&file->f_path); 2505 } 2506 2507 extern int file_modified(struct file *file); 2508 2509 int sync_inode_metadata(struct inode *inode, int wait); 2510 2511 struct file_system_type { 2512 const char *name; 2513 int fs_flags; 2514 #define FS_REQUIRES_DEV 1 2515 #define FS_BINARY_MOUNTDATA 2 2516 #define FS_HAS_SUBTYPE 4 2517 #define FS_USERNS_MOUNT 8 /* Can be mounted by userns root */ 2518 #define FS_DISALLOW_NOTIFY_PERM 16 /* Disable fanotify permission events */ 2519 #define FS_ALLOW_IDMAP 32 /* FS has been updated to handle vfs idmappings. */ 2520 #define FS_THP_SUPPORT 8192 /* Remove once all fs converted */ 2521 #define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() during rename() internally. */ 2522 int (*init_fs_context)(struct fs_context *); 2523 const struct fs_parameter_spec *parameters; 2524 struct dentry *(*mount) (struct file_system_type *, int, 2525 const char *, void *); 2526 void (*kill_sb) (struct super_block *); 2527 struct module *owner; 2528 struct file_system_type * next; 2529 struct hlist_head fs_supers; 2530 2531 struct lock_class_key s_lock_key; 2532 struct lock_class_key s_umount_key; 2533 struct lock_class_key s_vfs_rename_key; 2534 struct lock_class_key s_writers_key[SB_FREEZE_LEVELS]; 2535 2536 struct lock_class_key i_lock_key; 2537 struct lock_class_key i_mutex_key; 2538 struct lock_class_key invalidate_lock_key; 2539 struct lock_class_key i_mutex_dir_key; 2540 }; 2541 2542 #define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME) 2543 2544 extern struct dentry *mount_bdev(struct file_system_type *fs_type, 2545 int flags, const char *dev_name, void *data, 2546 int (*fill_super)(struct super_block *, void *, int)); 2547 extern struct dentry *mount_single(struct file_system_type *fs_type, 2548 int flags, void *data, 2549 int (*fill_super)(struct super_block *, void *, int)); 2550 extern struct dentry *mount_nodev(struct file_system_type *fs_type, 2551 int flags, void *data, 2552 int (*fill_super)(struct super_block *, void *, int)); 2553 extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path); 2554 void generic_shutdown_super(struct super_block *sb); 2555 void kill_block_super(struct super_block *sb); 2556 void kill_anon_super(struct super_block *sb); 2557 void kill_litter_super(struct super_block *sb); 2558 void deactivate_super(struct super_block *sb); 2559 void deactivate_locked_super(struct super_block *sb); 2560 int set_anon_super(struct super_block *s, void *data); 2561 int set_anon_super_fc(struct super_block *s, struct fs_context *fc); 2562 int get_anon_bdev(dev_t *); 2563 void free_anon_bdev(dev_t); 2564 struct super_block *sget_fc(struct fs_context *fc, 2565 int (*test)(struct super_block *, struct fs_context *), 2566 int (*set)(struct super_block *, struct fs_context *)); 2567 struct super_block *sget(struct file_system_type *type, 2568 int (*test)(struct super_block *,void *), 2569 int (*set)(struct super_block *,void *), 2570 int flags, void *data); 2571 2572 /* Alas, no aliases. Too much hassle with bringing module.h everywhere */ 2573 #define fops_get(fops) \ 2574 (((fops) && try_module_get((fops)->owner) ? (fops) : NULL)) 2575 #define fops_put(fops) \ 2576 do { if (fops) module_put((fops)->owner); } while(0) 2577 /* 2578 * This one is to be used *ONLY* from ->open() instances. 2579 * fops must be non-NULL, pinned down *and* module dependencies 2580 * should be sufficient to pin the caller down as well. 2581 */ 2582 #define replace_fops(f, fops) \ 2583 do { \ 2584 struct file *__file = (f); \ 2585 fops_put(__file->f_op); \ 2586 BUG_ON(!(__file->f_op = (fops))); \ 2587 } while(0) 2588 2589 extern int register_filesystem(struct file_system_type *); 2590 extern int unregister_filesystem(struct file_system_type *); 2591 extern struct vfsmount *kern_mount(struct file_system_type *); 2592 extern void kern_unmount(struct vfsmount *mnt); 2593 extern int may_umount_tree(struct vfsmount *); 2594 extern int may_umount(struct vfsmount *); 2595 extern long do_mount(const char *, const char __user *, 2596 const char *, unsigned long, void *); 2597 extern struct vfsmount *collect_mounts(const struct path *); 2598 extern void drop_collected_mounts(struct vfsmount *); 2599 extern int iterate_mounts(int (*)(struct vfsmount *, void *), void *, 2600 struct vfsmount *); 2601 extern int vfs_statfs(const struct path *, struct kstatfs *); 2602 extern int user_statfs(const char __user *, struct kstatfs *); 2603 extern int fd_statfs(int, struct kstatfs *); 2604 extern int freeze_super(struct super_block *super); 2605 extern int thaw_super(struct super_block *super); 2606 extern bool our_mnt(struct vfsmount *mnt); 2607 extern __printf(2, 3) 2608 int super_setup_bdi_name(struct super_block *sb, char *fmt, ...); 2609 extern int super_setup_bdi(struct super_block *sb); 2610 2611 extern int current_umask(void); 2612 2613 extern void ihold(struct inode * inode); 2614 extern void iput(struct inode *); 2615 extern int generic_update_time(struct inode *, struct timespec64 *, int); 2616 2617 /* /sys/fs */ 2618 extern struct kobject *fs_kobj; 2619 2620 #define MAX_RW_COUNT (INT_MAX & PAGE_MASK) 2621 2622 #ifdef CONFIG_FILE_LOCKING 2623 static inline int break_lease(struct inode *inode, unsigned int mode) 2624 { 2625 /* 2626 * Since this check is lockless, we must ensure that any refcounts 2627 * taken are done before checking i_flctx->flc_lease. Otherwise, we 2628 * could end up racing with tasks trying to set a new lease on this 2629 * file. 2630 */ 2631 smp_mb(); 2632 if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease)) 2633 return __break_lease(inode, mode, FL_LEASE); 2634 return 0; 2635 } 2636 2637 static inline int break_deleg(struct inode *inode, unsigned int mode) 2638 { 2639 /* 2640 * Since this check is lockless, we must ensure that any refcounts 2641 * taken are done before checking i_flctx->flc_lease. Otherwise, we 2642 * could end up racing with tasks trying to set a new lease on this 2643 * file. 2644 */ 2645 smp_mb(); 2646 if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease)) 2647 return __break_lease(inode, mode, FL_DELEG); 2648 return 0; 2649 } 2650 2651 static inline int try_break_deleg(struct inode *inode, struct inode **delegated_inode) 2652 { 2653 int ret; 2654 2655 ret = break_deleg(inode, O_WRONLY|O_NONBLOCK); 2656 if (ret == -EWOULDBLOCK && delegated_inode) { 2657 *delegated_inode = inode; 2658 ihold(inode); 2659 } 2660 return ret; 2661 } 2662 2663 static inline int break_deleg_wait(struct inode **delegated_inode) 2664 { 2665 int ret; 2666 2667 ret = break_deleg(*delegated_inode, O_WRONLY); 2668 iput(*delegated_inode); 2669 *delegated_inode = NULL; 2670 return ret; 2671 } 2672 2673 static inline int break_layout(struct inode *inode, bool wait) 2674 { 2675 smp_mb(); 2676 if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease)) 2677 return __break_lease(inode, 2678 wait ? O_WRONLY : O_WRONLY | O_NONBLOCK, 2679 FL_LAYOUT); 2680 return 0; 2681 } 2682 2683 #else /* !CONFIG_FILE_LOCKING */ 2684 static inline int break_lease(struct inode *inode, unsigned int mode) 2685 { 2686 return 0; 2687 } 2688 2689 static inline int break_deleg(struct inode *inode, unsigned int mode) 2690 { 2691 return 0; 2692 } 2693 2694 static inline int try_break_deleg(struct inode *inode, struct inode **delegated_inode) 2695 { 2696 return 0; 2697 } 2698 2699 static inline int break_deleg_wait(struct inode **delegated_inode) 2700 { 2701 BUG(); 2702 return 0; 2703 } 2704 2705 static inline int break_layout(struct inode *inode, bool wait) 2706 { 2707 return 0; 2708 } 2709 2710 #endif /* CONFIG_FILE_LOCKING */ 2711 2712 /* fs/open.c */ 2713 struct audit_names; 2714 struct filename { 2715 const char *name; /* pointer to actual string */ 2716 const __user char *uptr; /* original userland pointer */ 2717 int refcnt; 2718 struct audit_names *aname; 2719 const char iname[]; 2720 }; 2721 static_assert(offsetof(struct filename, iname) % sizeof(long) == 0); 2722 2723 static inline struct user_namespace *file_mnt_user_ns(struct file *file) 2724 { 2725 return mnt_user_ns(file->f_path.mnt); 2726 } 2727 extern long vfs_truncate(const struct path *, loff_t); 2728 int do_truncate(struct user_namespace *, struct dentry *, loff_t start, 2729 unsigned int time_attrs, struct file *filp); 2730 extern int vfs_fallocate(struct file *file, int mode, loff_t offset, 2731 loff_t len); 2732 extern long do_sys_open(int dfd, const char __user *filename, int flags, 2733 umode_t mode); 2734 extern struct file *file_open_name(struct filename *, int, umode_t); 2735 extern struct file *filp_open(const char *, int, umode_t); 2736 extern struct file *file_open_root(const struct path *, 2737 const char *, int, umode_t); 2738 static inline struct file *file_open_root_mnt(struct vfsmount *mnt, 2739 const char *name, int flags, umode_t mode) 2740 { 2741 return file_open_root(&(struct path){.mnt = mnt, .dentry = mnt->mnt_root}, 2742 name, flags, mode); 2743 } 2744 extern struct file * dentry_open(const struct path *, int, const struct cred *); 2745 extern struct file * open_with_fake_path(const struct path *, int, 2746 struct inode*, const struct cred *); 2747 static inline struct file *file_clone_open(struct file *file) 2748 { 2749 return dentry_open(&file->f_path, file->f_flags, file->f_cred); 2750 } 2751 extern int filp_close(struct file *, fl_owner_t id); 2752 2753 extern struct filename *getname_flags(const char __user *, int, int *); 2754 extern struct filename *getname_uflags(const char __user *, int); 2755 extern struct filename *getname(const char __user *); 2756 extern struct filename *getname_kernel(const char *); 2757 extern void putname(struct filename *name); 2758 2759 extern int finish_open(struct file *file, struct dentry *dentry, 2760 int (*open)(struct inode *, struct file *)); 2761 extern int finish_no_open(struct file *file, struct dentry *dentry); 2762 2763 /* fs/dcache.c */ 2764 extern void __init vfs_caches_init_early(void); 2765 extern void __init vfs_caches_init(void); 2766 2767 extern struct kmem_cache *names_cachep; 2768 2769 #define __getname() kmem_cache_alloc(names_cachep, GFP_KERNEL) 2770 #define __putname(name) kmem_cache_free(names_cachep, (void *)(name)) 2771 2772 extern struct super_block *blockdev_superblock; 2773 static inline bool sb_is_blkdev_sb(struct super_block *sb) 2774 { 2775 return IS_ENABLED(CONFIG_BLOCK) && sb == blockdev_superblock; 2776 } 2777 2778 void emergency_thaw_all(void); 2779 extern int sync_filesystem(struct super_block *); 2780 extern const struct file_operations def_blk_fops; 2781 extern const struct file_operations def_chr_fops; 2782 2783 /* fs/char_dev.c */ 2784 #define CHRDEV_MAJOR_MAX 512 2785 /* Marks the bottom of the first segment of free char majors */ 2786 #define CHRDEV_MAJOR_DYN_END 234 2787 /* Marks the top and bottom of the second segment of free char majors */ 2788 #define CHRDEV_MAJOR_DYN_EXT_START 511 2789 #define CHRDEV_MAJOR_DYN_EXT_END 384 2790 2791 extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *); 2792 extern int register_chrdev_region(dev_t, unsigned, const char *); 2793 extern int __register_chrdev(unsigned int major, unsigned int baseminor, 2794 unsigned int count, const char *name, 2795 const struct file_operations *fops); 2796 extern void __unregister_chrdev(unsigned int major, unsigned int baseminor, 2797 unsigned int count, const char *name); 2798 extern void unregister_chrdev_region(dev_t, unsigned); 2799 extern void chrdev_show(struct seq_file *,off_t); 2800 2801 static inline int register_chrdev(unsigned int major, const char *name, 2802 const struct file_operations *fops) 2803 { 2804 return __register_chrdev(major, 0, 256, name, fops); 2805 } 2806 2807 static inline void unregister_chrdev(unsigned int major, const char *name) 2808 { 2809 __unregister_chrdev(major, 0, 256, name); 2810 } 2811 2812 extern void init_special_inode(struct inode *, umode_t, dev_t); 2813 2814 /* Invalid inode operations -- fs/bad_inode.c */ 2815 extern void make_bad_inode(struct inode *); 2816 extern bool is_bad_inode(struct inode *); 2817 2818 unsigned long invalidate_mapping_pages(struct address_space *mapping, 2819 pgoff_t start, pgoff_t end); 2820 2821 void invalidate_mapping_pagevec(struct address_space *mapping, 2822 pgoff_t start, pgoff_t end, 2823 unsigned long *nr_pagevec); 2824 2825 static inline void invalidate_remote_inode(struct inode *inode) 2826 { 2827 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 2828 S_ISLNK(inode->i_mode)) 2829 invalidate_mapping_pages(inode->i_mapping, 0, -1); 2830 } 2831 extern int invalidate_inode_pages2(struct address_space *mapping); 2832 extern int invalidate_inode_pages2_range(struct address_space *mapping, 2833 pgoff_t start, pgoff_t end); 2834 extern int write_inode_now(struct inode *, int); 2835 extern int filemap_fdatawrite(struct address_space *); 2836 extern int filemap_flush(struct address_space *); 2837 extern int filemap_fdatawait_keep_errors(struct address_space *mapping); 2838 extern int filemap_fdatawait_range(struct address_space *, loff_t lstart, 2839 loff_t lend); 2840 extern int filemap_fdatawait_range_keep_errors(struct address_space *mapping, 2841 loff_t start_byte, loff_t end_byte); 2842 2843 static inline int filemap_fdatawait(struct address_space *mapping) 2844 { 2845 return filemap_fdatawait_range(mapping, 0, LLONG_MAX); 2846 } 2847 2848 extern bool filemap_range_has_page(struct address_space *, loff_t lstart, 2849 loff_t lend); 2850 extern bool filemap_range_needs_writeback(struct address_space *, 2851 loff_t lstart, loff_t lend); 2852 extern int filemap_write_and_wait_range(struct address_space *mapping, 2853 loff_t lstart, loff_t lend); 2854 extern int __filemap_fdatawrite_range(struct address_space *mapping, 2855 loff_t start, loff_t end, int sync_mode); 2856 extern int filemap_fdatawrite_range(struct address_space *mapping, 2857 loff_t start, loff_t end); 2858 extern int filemap_check_errors(struct address_space *mapping); 2859 extern void __filemap_set_wb_err(struct address_space *mapping, int err); 2860 int filemap_fdatawrite_wbc(struct address_space *mapping, 2861 struct writeback_control *wbc); 2862 2863 static inline int filemap_write_and_wait(struct address_space *mapping) 2864 { 2865 return filemap_write_and_wait_range(mapping, 0, LLONG_MAX); 2866 } 2867 2868 extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart, 2869 loff_t lend); 2870 extern int __must_check file_check_and_advance_wb_err(struct file *file); 2871 extern int __must_check file_write_and_wait_range(struct file *file, 2872 loff_t start, loff_t end); 2873 2874 static inline int file_write_and_wait(struct file *file) 2875 { 2876 return file_write_and_wait_range(file, 0, LLONG_MAX); 2877 } 2878 2879 /** 2880 * filemap_set_wb_err - set a writeback error on an address_space 2881 * @mapping: mapping in which to set writeback error 2882 * @err: error to be set in mapping 2883 * 2884 * When writeback fails in some way, we must record that error so that 2885 * userspace can be informed when fsync and the like are called. We endeavor 2886 * to report errors on any file that was open at the time of the error. Some 2887 * internal callers also need to know when writeback errors have occurred. 2888 * 2889 * When a writeback error occurs, most filesystems will want to call 2890 * filemap_set_wb_err to record the error in the mapping so that it will be 2891 * automatically reported whenever fsync is called on the file. 2892 */ 2893 static inline void filemap_set_wb_err(struct address_space *mapping, int err) 2894 { 2895 /* Fastpath for common case of no error */ 2896 if (unlikely(err)) 2897 __filemap_set_wb_err(mapping, err); 2898 } 2899 2900 /** 2901 * filemap_check_wb_err - has an error occurred since the mark was sampled? 2902 * @mapping: mapping to check for writeback errors 2903 * @since: previously-sampled errseq_t 2904 * 2905 * Grab the errseq_t value from the mapping, and see if it has changed "since" 2906 * the given value was sampled. 2907 * 2908 * If it has then report the latest error set, otherwise return 0. 2909 */ 2910 static inline int filemap_check_wb_err(struct address_space *mapping, 2911 errseq_t since) 2912 { 2913 return errseq_check(&mapping->wb_err, since); 2914 } 2915 2916 /** 2917 * filemap_sample_wb_err - sample the current errseq_t to test for later errors 2918 * @mapping: mapping to be sampled 2919 * 2920 * Writeback errors are always reported relative to a particular sample point 2921 * in the past. This function provides those sample points. 2922 */ 2923 static inline errseq_t filemap_sample_wb_err(struct address_space *mapping) 2924 { 2925 return errseq_sample(&mapping->wb_err); 2926 } 2927 2928 /** 2929 * file_sample_sb_err - sample the current errseq_t to test for later errors 2930 * @file: file pointer to be sampled 2931 * 2932 * Grab the most current superblock-level errseq_t value for the given 2933 * struct file. 2934 */ 2935 static inline errseq_t file_sample_sb_err(struct file *file) 2936 { 2937 return errseq_sample(&file->f_path.dentry->d_sb->s_wb_err); 2938 } 2939 2940 extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end, 2941 int datasync); 2942 extern int vfs_fsync(struct file *file, int datasync); 2943 2944 extern int sync_file_range(struct file *file, loff_t offset, loff_t nbytes, 2945 unsigned int flags); 2946 2947 /* 2948 * Sync the bytes written if this was a synchronous write. Expect ki_pos 2949 * to already be updated for the write, and will return either the amount 2950 * of bytes passed in, or an error if syncing the file failed. 2951 */ 2952 static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count) 2953 { 2954 if (iocb->ki_flags & IOCB_DSYNC) { 2955 int ret = vfs_fsync_range(iocb->ki_filp, 2956 iocb->ki_pos - count, iocb->ki_pos - 1, 2957 (iocb->ki_flags & IOCB_SYNC) ? 0 : 1); 2958 if (ret) 2959 return ret; 2960 } 2961 2962 return count; 2963 } 2964 2965 extern void emergency_sync(void); 2966 extern void emergency_remount(void); 2967 2968 #ifdef CONFIG_BLOCK 2969 extern int bmap(struct inode *inode, sector_t *block); 2970 #else 2971 static inline int bmap(struct inode *inode, sector_t *block) 2972 { 2973 return -EINVAL; 2974 } 2975 #endif 2976 2977 int notify_change(struct user_namespace *, struct dentry *, 2978 struct iattr *, struct inode **); 2979 int inode_permission(struct user_namespace *, struct inode *, int); 2980 int generic_permission(struct user_namespace *, struct inode *, int); 2981 static inline int file_permission(struct file *file, int mask) 2982 { 2983 return inode_permission(file_mnt_user_ns(file), 2984 file_inode(file), mask); 2985 } 2986 static inline int path_permission(const struct path *path, int mask) 2987 { 2988 return inode_permission(mnt_user_ns(path->mnt), 2989 d_inode(path->dentry), mask); 2990 } 2991 int __check_sticky(struct user_namespace *mnt_userns, struct inode *dir, 2992 struct inode *inode); 2993 2994 static inline bool execute_ok(struct inode *inode) 2995 { 2996 return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode); 2997 } 2998 2999 static inline bool inode_wrong_type(const struct inode *inode, umode_t mode) 3000 { 3001 return (inode->i_mode ^ mode) & S_IFMT; 3002 } 3003 3004 static inline void file_start_write(struct file *file) 3005 { 3006 if (!S_ISREG(file_inode(file)->i_mode)) 3007 return; 3008 sb_start_write(file_inode(file)->i_sb); 3009 } 3010 3011 static inline bool file_start_write_trylock(struct file *file) 3012 { 3013 if (!S_ISREG(file_inode(file)->i_mode)) 3014 return true; 3015 return sb_start_write_trylock(file_inode(file)->i_sb); 3016 } 3017 3018 static inline void file_end_write(struct file *file) 3019 { 3020 if (!S_ISREG(file_inode(file)->i_mode)) 3021 return; 3022 __sb_end_write(file_inode(file)->i_sb, SB_FREEZE_WRITE); 3023 } 3024 3025 /* 3026 * This is used for regular files where some users -- especially the 3027 * currently executed binary in a process, previously handled via 3028 * VM_DENYWRITE -- cannot handle concurrent write (and maybe mmap 3029 * read-write shared) accesses. 3030 * 3031 * get_write_access() gets write permission for a file. 3032 * put_write_access() releases this write permission. 3033 * deny_write_access() denies write access to a file. 3034 * allow_write_access() re-enables write access to a file. 3035 * 3036 * The i_writecount field of an inode can have the following values: 3037 * 0: no write access, no denied write access 3038 * < 0: (-i_writecount) users that denied write access to the file. 3039 * > 0: (i_writecount) users that have write access to the file. 3040 * 3041 * Normally we operate on that counter with atomic_{inc,dec} and it's safe 3042 * except for the cases where we don't hold i_writecount yet. Then we need to 3043 * use {get,deny}_write_access() - these functions check the sign and refuse 3044 * to do the change if sign is wrong. 3045 */ 3046 static inline int get_write_access(struct inode *inode) 3047 { 3048 return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY; 3049 } 3050 static inline int deny_write_access(struct file *file) 3051 { 3052 struct inode *inode = file_inode(file); 3053 return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY; 3054 } 3055 static inline void put_write_access(struct inode * inode) 3056 { 3057 atomic_dec(&inode->i_writecount); 3058 } 3059 static inline void allow_write_access(struct file *file) 3060 { 3061 if (file) 3062 atomic_inc(&file_inode(file)->i_writecount); 3063 } 3064 static inline bool inode_is_open_for_write(const struct inode *inode) 3065 { 3066 return atomic_read(&inode->i_writecount) > 0; 3067 } 3068 3069 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING) 3070 static inline void i_readcount_dec(struct inode *inode) 3071 { 3072 BUG_ON(!atomic_read(&inode->i_readcount)); 3073 atomic_dec(&inode->i_readcount); 3074 } 3075 static inline void i_readcount_inc(struct inode *inode) 3076 { 3077 atomic_inc(&inode->i_readcount); 3078 } 3079 #else 3080 static inline void i_readcount_dec(struct inode *inode) 3081 { 3082 return; 3083 } 3084 static inline void i_readcount_inc(struct inode *inode) 3085 { 3086 return; 3087 } 3088 #endif 3089 extern int do_pipe_flags(int *, int); 3090 3091 extern ssize_t kernel_read(struct file *, void *, size_t, loff_t *); 3092 ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos); 3093 extern ssize_t kernel_write(struct file *, const void *, size_t, loff_t *); 3094 extern ssize_t __kernel_write(struct file *, const void *, size_t, loff_t *); 3095 extern struct file * open_exec(const char *); 3096 3097 /* fs/dcache.c -- generic fs support functions */ 3098 extern bool is_subdir(struct dentry *, struct dentry *); 3099 extern bool path_is_under(const struct path *, const struct path *); 3100 3101 extern char *file_path(struct file *, char *, int); 3102 3103 #include <linux/err.h> 3104 3105 /* needed for stackable file system support */ 3106 extern loff_t default_llseek(struct file *file, loff_t offset, int whence); 3107 3108 extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence); 3109 3110 extern int inode_init_always(struct super_block *, struct inode *); 3111 extern void inode_init_once(struct inode *); 3112 extern void address_space_init_once(struct address_space *mapping); 3113 extern struct inode * igrab(struct inode *); 3114 extern ino_t iunique(struct super_block *, ino_t); 3115 extern int inode_needs_sync(struct inode *inode); 3116 extern int generic_delete_inode(struct inode *inode); 3117 static inline int generic_drop_inode(struct inode *inode) 3118 { 3119 return !inode->i_nlink || inode_unhashed(inode); 3120 } 3121 extern void d_mark_dontcache(struct inode *inode); 3122 3123 extern struct inode *ilookup5_nowait(struct super_block *sb, 3124 unsigned long hashval, int (*test)(struct inode *, void *), 3125 void *data); 3126 extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval, 3127 int (*test)(struct inode *, void *), void *data); 3128 extern struct inode *ilookup(struct super_block *sb, unsigned long ino); 3129 3130 extern struct inode *inode_insert5(struct inode *inode, unsigned long hashval, 3131 int (*test)(struct inode *, void *), 3132 int (*set)(struct inode *, void *), 3133 void *data); 3134 extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *); 3135 extern struct inode * iget_locked(struct super_block *, unsigned long); 3136 extern struct inode *find_inode_nowait(struct super_block *, 3137 unsigned long, 3138 int (*match)(struct inode *, 3139 unsigned long, void *), 3140 void *data); 3141 extern struct inode *find_inode_rcu(struct super_block *, unsigned long, 3142 int (*)(struct inode *, void *), void *); 3143 extern struct inode *find_inode_by_ino_rcu(struct super_block *, unsigned long); 3144 extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *); 3145 extern int insert_inode_locked(struct inode *); 3146 #ifdef CONFIG_DEBUG_LOCK_ALLOC 3147 extern void lockdep_annotate_inode_mutex_key(struct inode *inode); 3148 #else 3149 static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { }; 3150 #endif 3151 extern void unlock_new_inode(struct inode *); 3152 extern void discard_new_inode(struct inode *); 3153 extern unsigned int get_next_ino(void); 3154 extern void evict_inodes(struct super_block *sb); 3155 3156 /* 3157 * Userspace may rely on the the inode number being non-zero. For example, glibc 3158 * simply ignores files with zero i_ino in unlink() and other places. 3159 * 3160 * As an additional complication, if userspace was compiled with 3161 * _FILE_OFFSET_BITS=32 on a 64-bit kernel we'll only end up reading out the 3162 * lower 32 bits, so we need to check that those aren't zero explicitly. With 3163 * _FILE_OFFSET_BITS=64, this may cause some harmless false-negatives, but 3164 * better safe than sorry. 3165 */ 3166 static inline bool is_zero_ino(ino_t ino) 3167 { 3168 return (u32)ino == 0; 3169 } 3170 3171 extern void __iget(struct inode * inode); 3172 extern void iget_failed(struct inode *); 3173 extern void clear_inode(struct inode *); 3174 extern void __destroy_inode(struct inode *); 3175 extern struct inode *new_inode_pseudo(struct super_block *sb); 3176 extern struct inode *new_inode(struct super_block *sb); 3177 extern void free_inode_nonrcu(struct inode *inode); 3178 extern int should_remove_suid(struct dentry *); 3179 extern int file_remove_privs(struct file *); 3180 3181 extern void __insert_inode_hash(struct inode *, unsigned long hashval); 3182 static inline void insert_inode_hash(struct inode *inode) 3183 { 3184 __insert_inode_hash(inode, inode->i_ino); 3185 } 3186 3187 extern void __remove_inode_hash(struct inode *); 3188 static inline void remove_inode_hash(struct inode *inode) 3189 { 3190 if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash)) 3191 __remove_inode_hash(inode); 3192 } 3193 3194 extern void inode_sb_list_add(struct inode *inode); 3195 3196 extern int sb_set_blocksize(struct super_block *, int); 3197 extern int sb_min_blocksize(struct super_block *, int); 3198 3199 extern int generic_file_mmap(struct file *, struct vm_area_struct *); 3200 extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *); 3201 extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *); 3202 extern int generic_write_check_limits(struct file *file, loff_t pos, 3203 loff_t *count); 3204 extern int generic_file_rw_checks(struct file *file_in, struct file *file_out); 3205 ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *to, 3206 ssize_t already_read); 3207 extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *); 3208 extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *); 3209 extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *); 3210 extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *); 3211 extern ssize_t generic_perform_write(struct file *, struct iov_iter *, loff_t); 3212 3213 ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos, 3214 rwf_t flags); 3215 ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos, 3216 rwf_t flags); 3217 ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb, 3218 struct iov_iter *iter); 3219 ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb, 3220 struct iov_iter *iter); 3221 3222 /* fs/splice.c */ 3223 extern ssize_t generic_file_splice_read(struct file *, loff_t *, 3224 struct pipe_inode_info *, size_t, unsigned int); 3225 extern ssize_t iter_file_splice_write(struct pipe_inode_info *, 3226 struct file *, loff_t *, size_t, unsigned int); 3227 extern ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, 3228 struct file *out, loff_t *, size_t len, unsigned int flags); 3229 extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out, 3230 loff_t *opos, size_t len, unsigned int flags); 3231 3232 3233 extern void 3234 file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping); 3235 extern loff_t noop_llseek(struct file *file, loff_t offset, int whence); 3236 extern loff_t no_llseek(struct file *file, loff_t offset, int whence); 3237 extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize); 3238 extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence); 3239 extern loff_t generic_file_llseek_size(struct file *file, loff_t offset, 3240 int whence, loff_t maxsize, loff_t eof); 3241 extern loff_t fixed_size_llseek(struct file *file, loff_t offset, 3242 int whence, loff_t size); 3243 extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t); 3244 extern loff_t no_seek_end_llseek(struct file *, loff_t, int); 3245 extern int generic_file_open(struct inode * inode, struct file * filp); 3246 extern int nonseekable_open(struct inode * inode, struct file * filp); 3247 extern int stream_open(struct inode * inode, struct file * filp); 3248 3249 #ifdef CONFIG_BLOCK 3250 typedef void (dio_submit_t)(struct bio *bio, struct inode *inode, 3251 loff_t file_offset); 3252 3253 enum { 3254 /* need locking between buffered and direct access */ 3255 DIO_LOCKING = 0x01, 3256 3257 /* filesystem does not support filling holes */ 3258 DIO_SKIP_HOLES = 0x02, 3259 }; 3260 3261 ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode, 3262 struct block_device *bdev, struct iov_iter *iter, 3263 get_block_t get_block, 3264 dio_iodone_t end_io, dio_submit_t submit_io, 3265 int flags); 3266 3267 static inline ssize_t blockdev_direct_IO(struct kiocb *iocb, 3268 struct inode *inode, 3269 struct iov_iter *iter, 3270 get_block_t get_block) 3271 { 3272 return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter, 3273 get_block, NULL, NULL, DIO_LOCKING | DIO_SKIP_HOLES); 3274 } 3275 #endif 3276 3277 void inode_dio_wait(struct inode *inode); 3278 3279 /** 3280 * inode_dio_begin - signal start of a direct I/O requests 3281 * @inode: inode the direct I/O happens on 3282 * 3283 * This is called once we've finished processing a direct I/O request, 3284 * and is used to wake up callers waiting for direct I/O to be quiesced. 3285 */ 3286 static inline void inode_dio_begin(struct inode *inode) 3287 { 3288 atomic_inc(&inode->i_dio_count); 3289 } 3290 3291 /** 3292 * inode_dio_end - signal finish of a direct I/O requests 3293 * @inode: inode the direct I/O happens on 3294 * 3295 * This is called once we've finished processing a direct I/O request, 3296 * and is used to wake up callers waiting for direct I/O to be quiesced. 3297 */ 3298 static inline void inode_dio_end(struct inode *inode) 3299 { 3300 if (atomic_dec_and_test(&inode->i_dio_count)) 3301 wake_up_bit(&inode->i_state, __I_DIO_WAKEUP); 3302 } 3303 3304 /* 3305 * Warn about a page cache invalidation failure diring a direct I/O write. 3306 */ 3307 void dio_warn_stale_pagecache(struct file *filp); 3308 3309 extern void inode_set_flags(struct inode *inode, unsigned int flags, 3310 unsigned int mask); 3311 3312 extern const struct file_operations generic_ro_fops; 3313 3314 #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m)) 3315 3316 extern int readlink_copy(char __user *, int, const char *); 3317 extern int page_readlink(struct dentry *, char __user *, int); 3318 extern const char *page_get_link(struct dentry *, struct inode *, 3319 struct delayed_call *); 3320 extern void page_put_link(void *); 3321 extern int __page_symlink(struct inode *inode, const char *symname, int len, 3322 int nofs); 3323 extern int page_symlink(struct inode *inode, const char *symname, int len); 3324 extern const struct inode_operations page_symlink_inode_operations; 3325 extern void kfree_link(void *); 3326 void generic_fillattr(struct user_namespace *, struct inode *, struct kstat *); 3327 void generic_fill_statx_attr(struct inode *inode, struct kstat *stat); 3328 extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int); 3329 extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int); 3330 void __inode_add_bytes(struct inode *inode, loff_t bytes); 3331 void inode_add_bytes(struct inode *inode, loff_t bytes); 3332 void __inode_sub_bytes(struct inode *inode, loff_t bytes); 3333 void inode_sub_bytes(struct inode *inode, loff_t bytes); 3334 static inline loff_t __inode_get_bytes(struct inode *inode) 3335 { 3336 return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes; 3337 } 3338 loff_t inode_get_bytes(struct inode *inode); 3339 void inode_set_bytes(struct inode *inode, loff_t bytes); 3340 const char *simple_get_link(struct dentry *, struct inode *, 3341 struct delayed_call *); 3342 extern const struct inode_operations simple_symlink_inode_operations; 3343 3344 extern int iterate_dir(struct file *, struct dir_context *); 3345 3346 int vfs_fstatat(int dfd, const char __user *filename, struct kstat *stat, 3347 int flags); 3348 int vfs_fstat(int fd, struct kstat *stat); 3349 3350 static inline int vfs_stat(const char __user *filename, struct kstat *stat) 3351 { 3352 return vfs_fstatat(AT_FDCWD, filename, stat, 0); 3353 } 3354 static inline int vfs_lstat(const char __user *name, struct kstat *stat) 3355 { 3356 return vfs_fstatat(AT_FDCWD, name, stat, AT_SYMLINK_NOFOLLOW); 3357 } 3358 3359 extern const char *vfs_get_link(struct dentry *, struct delayed_call *); 3360 extern int vfs_readlink(struct dentry *, char __user *, int); 3361 3362 extern struct file_system_type *get_filesystem(struct file_system_type *fs); 3363 extern void put_filesystem(struct file_system_type *fs); 3364 extern struct file_system_type *get_fs_type(const char *name); 3365 extern struct super_block *get_super(struct block_device *); 3366 extern struct super_block *get_active_super(struct block_device *bdev); 3367 extern void drop_super(struct super_block *sb); 3368 extern void drop_super_exclusive(struct super_block *sb); 3369 extern void iterate_supers(void (*)(struct super_block *, void *), void *); 3370 extern void iterate_supers_type(struct file_system_type *, 3371 void (*)(struct super_block *, void *), void *); 3372 3373 extern int dcache_dir_open(struct inode *, struct file *); 3374 extern int dcache_dir_close(struct inode *, struct file *); 3375 extern loff_t dcache_dir_lseek(struct file *, loff_t, int); 3376 extern int dcache_readdir(struct file *, struct dir_context *); 3377 extern int simple_setattr(struct user_namespace *, struct dentry *, 3378 struct iattr *); 3379 extern int simple_getattr(struct user_namespace *, const struct path *, 3380 struct kstat *, u32, unsigned int); 3381 extern int simple_statfs(struct dentry *, struct kstatfs *); 3382 extern int simple_open(struct inode *inode, struct file *file); 3383 extern int simple_link(struct dentry *, struct inode *, struct dentry *); 3384 extern int simple_unlink(struct inode *, struct dentry *); 3385 extern int simple_rmdir(struct inode *, struct dentry *); 3386 extern int simple_rename(struct user_namespace *, struct inode *, 3387 struct dentry *, struct inode *, struct dentry *, 3388 unsigned int); 3389 extern void simple_recursive_removal(struct dentry *, 3390 void (*callback)(struct dentry *)); 3391 extern int noop_fsync(struct file *, loff_t, loff_t, int); 3392 extern void noop_invalidatepage(struct page *page, unsigned int offset, 3393 unsigned int length); 3394 extern ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter); 3395 extern int simple_empty(struct dentry *); 3396 extern int simple_write_begin(struct file *file, struct address_space *mapping, 3397 loff_t pos, unsigned len, unsigned flags, 3398 struct page **pagep, void **fsdata); 3399 extern const struct address_space_operations ram_aops; 3400 extern int always_delete_dentry(const struct dentry *); 3401 extern struct inode *alloc_anon_inode(struct super_block *); 3402 extern int simple_nosetlease(struct file *, long, struct file_lock **, void **); 3403 extern const struct dentry_operations simple_dentry_operations; 3404 3405 extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags); 3406 extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *); 3407 extern const struct file_operations simple_dir_operations; 3408 extern const struct inode_operations simple_dir_inode_operations; 3409 extern void make_empty_dir_inode(struct inode *inode); 3410 extern bool is_empty_dir_inode(struct inode *inode); 3411 struct tree_descr { const char *name; const struct file_operations *ops; int mode; }; 3412 struct dentry *d_alloc_name(struct dentry *, const char *); 3413 extern int simple_fill_super(struct super_block *, unsigned long, 3414 const struct tree_descr *); 3415 extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count); 3416 extern void simple_release_fs(struct vfsmount **mount, int *count); 3417 3418 extern ssize_t simple_read_from_buffer(void __user *to, size_t count, 3419 loff_t *ppos, const void *from, size_t available); 3420 extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos, 3421 const void __user *from, size_t count); 3422 3423 extern int __generic_file_fsync(struct file *, loff_t, loff_t, int); 3424 extern int generic_file_fsync(struct file *, loff_t, loff_t, int); 3425 3426 extern int generic_check_addressable(unsigned, u64); 3427 3428 extern void generic_set_encrypted_ci_d_ops(struct dentry *dentry); 3429 3430 #ifdef CONFIG_MIGRATION 3431 extern int buffer_migrate_page(struct address_space *, 3432 struct page *, struct page *, 3433 enum migrate_mode); 3434 extern int buffer_migrate_page_norefs(struct address_space *, 3435 struct page *, struct page *, 3436 enum migrate_mode); 3437 #else 3438 #define buffer_migrate_page NULL 3439 #define buffer_migrate_page_norefs NULL 3440 #endif 3441 3442 int may_setattr(struct user_namespace *mnt_userns, struct inode *inode, 3443 unsigned int ia_valid); 3444 int setattr_prepare(struct user_namespace *, struct dentry *, struct iattr *); 3445 extern int inode_newsize_ok(const struct inode *, loff_t offset); 3446 void setattr_copy(struct user_namespace *, struct inode *inode, 3447 const struct iattr *attr); 3448 3449 extern int file_update_time(struct file *file); 3450 3451 static inline bool vma_is_dax(const struct vm_area_struct *vma) 3452 { 3453 return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host); 3454 } 3455 3456 static inline bool vma_is_fsdax(struct vm_area_struct *vma) 3457 { 3458 struct inode *inode; 3459 3460 if (!IS_ENABLED(CONFIG_FS_DAX) || !vma->vm_file) 3461 return false; 3462 if (!vma_is_dax(vma)) 3463 return false; 3464 inode = file_inode(vma->vm_file); 3465 if (S_ISCHR(inode->i_mode)) 3466 return false; /* device-dax */ 3467 return true; 3468 } 3469 3470 static inline int iocb_flags(struct file *file) 3471 { 3472 int res = 0; 3473 if (file->f_flags & O_APPEND) 3474 res |= IOCB_APPEND; 3475 if (file->f_flags & O_DIRECT) 3476 res |= IOCB_DIRECT; 3477 if ((file->f_flags & O_DSYNC) || IS_SYNC(file->f_mapping->host)) 3478 res |= IOCB_DSYNC; 3479 if (file->f_flags & __O_SYNC) 3480 res |= IOCB_SYNC; 3481 return res; 3482 } 3483 3484 static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags) 3485 { 3486 int kiocb_flags = 0; 3487 3488 /* make sure there's no overlap between RWF and private IOCB flags */ 3489 BUILD_BUG_ON((__force int) RWF_SUPPORTED & IOCB_EVENTFD); 3490 3491 if (!flags) 3492 return 0; 3493 if (unlikely(flags & ~RWF_SUPPORTED)) 3494 return -EOPNOTSUPP; 3495 3496 if (flags & RWF_NOWAIT) { 3497 if (!(ki->ki_filp->f_mode & FMODE_NOWAIT)) 3498 return -EOPNOTSUPP; 3499 kiocb_flags |= IOCB_NOIO; 3500 } 3501 kiocb_flags |= (__force int) (flags & RWF_SUPPORTED); 3502 if (flags & RWF_SYNC) 3503 kiocb_flags |= IOCB_DSYNC; 3504 3505 ki->ki_flags |= kiocb_flags; 3506 return 0; 3507 } 3508 3509 static inline ino_t parent_ino(struct dentry *dentry) 3510 { 3511 ino_t res; 3512 3513 /* 3514 * Don't strictly need d_lock here? If the parent ino could change 3515 * then surely we'd have a deeper race in the caller? 3516 */ 3517 spin_lock(&dentry->d_lock); 3518 res = dentry->d_parent->d_inode->i_ino; 3519 spin_unlock(&dentry->d_lock); 3520 return res; 3521 } 3522 3523 /* Transaction based IO helpers */ 3524 3525 /* 3526 * An argresp is stored in an allocated page and holds the 3527 * size of the argument or response, along with its content 3528 */ 3529 struct simple_transaction_argresp { 3530 ssize_t size; 3531 char data[]; 3532 }; 3533 3534 #define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp)) 3535 3536 char *simple_transaction_get(struct file *file, const char __user *buf, 3537 size_t size); 3538 ssize_t simple_transaction_read(struct file *file, char __user *buf, 3539 size_t size, loff_t *pos); 3540 int simple_transaction_release(struct inode *inode, struct file *file); 3541 3542 void simple_transaction_set(struct file *file, size_t n); 3543 3544 /* 3545 * simple attribute files 3546 * 3547 * These attributes behave similar to those in sysfs: 3548 * 3549 * Writing to an attribute immediately sets a value, an open file can be 3550 * written to multiple times. 3551 * 3552 * Reading from an attribute creates a buffer from the value that might get 3553 * read with multiple read calls. When the attribute has been read 3554 * completely, no further read calls are possible until the file is opened 3555 * again. 3556 * 3557 * All attributes contain a text representation of a numeric value 3558 * that are accessed with the get() and set() functions. 3559 */ 3560 #define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt) \ 3561 static int __fops ## _open(struct inode *inode, struct file *file) \ 3562 { \ 3563 __simple_attr_check_format(__fmt, 0ull); \ 3564 return simple_attr_open(inode, file, __get, __set, __fmt); \ 3565 } \ 3566 static const struct file_operations __fops = { \ 3567 .owner = THIS_MODULE, \ 3568 .open = __fops ## _open, \ 3569 .release = simple_attr_release, \ 3570 .read = simple_attr_read, \ 3571 .write = simple_attr_write, \ 3572 .llseek = generic_file_llseek, \ 3573 } 3574 3575 static inline __printf(1, 2) 3576 void __simple_attr_check_format(const char *fmt, ...) 3577 { 3578 /* don't do anything, just let the compiler check the arguments; */ 3579 } 3580 3581 int simple_attr_open(struct inode *inode, struct file *file, 3582 int (*get)(void *, u64 *), int (*set)(void *, u64), 3583 const char *fmt); 3584 int simple_attr_release(struct inode *inode, struct file *file); 3585 ssize_t simple_attr_read(struct file *file, char __user *buf, 3586 size_t len, loff_t *ppos); 3587 ssize_t simple_attr_write(struct file *file, const char __user *buf, 3588 size_t len, loff_t *ppos); 3589 3590 struct ctl_table; 3591 int proc_nr_files(struct ctl_table *table, int write, 3592 void *buffer, size_t *lenp, loff_t *ppos); 3593 int proc_nr_dentry(struct ctl_table *table, int write, 3594 void *buffer, size_t *lenp, loff_t *ppos); 3595 int proc_nr_inodes(struct ctl_table *table, int write, 3596 void *buffer, size_t *lenp, loff_t *ppos); 3597 int __init list_bdev_fs_names(char *buf, size_t size); 3598 3599 #define __FMODE_EXEC ((__force int) FMODE_EXEC) 3600 #define __FMODE_NONOTIFY ((__force int) FMODE_NONOTIFY) 3601 3602 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE]) 3603 #define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \ 3604 (flag & __FMODE_NONOTIFY))) 3605 3606 static inline bool is_sxid(umode_t mode) 3607 { 3608 return (mode & S_ISUID) || ((mode & S_ISGID) && (mode & S_IXGRP)); 3609 } 3610 3611 static inline int check_sticky(struct user_namespace *mnt_userns, 3612 struct inode *dir, struct inode *inode) 3613 { 3614 if (!(dir->i_mode & S_ISVTX)) 3615 return 0; 3616 3617 return __check_sticky(mnt_userns, dir, inode); 3618 } 3619 3620 static inline void inode_has_no_xattr(struct inode *inode) 3621 { 3622 if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC)) 3623 inode->i_flags |= S_NOSEC; 3624 } 3625 3626 static inline bool is_root_inode(struct inode *inode) 3627 { 3628 return inode == inode->i_sb->s_root->d_inode; 3629 } 3630 3631 static inline bool dir_emit(struct dir_context *ctx, 3632 const char *name, int namelen, 3633 u64 ino, unsigned type) 3634 { 3635 return ctx->actor(ctx, name, namelen, ctx->pos, ino, type) == 0; 3636 } 3637 static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx) 3638 { 3639 return ctx->actor(ctx, ".", 1, ctx->pos, 3640 file->f_path.dentry->d_inode->i_ino, DT_DIR) == 0; 3641 } 3642 static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx) 3643 { 3644 return ctx->actor(ctx, "..", 2, ctx->pos, 3645 parent_ino(file->f_path.dentry), DT_DIR) == 0; 3646 } 3647 static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx) 3648 { 3649 if (ctx->pos == 0) { 3650 if (!dir_emit_dot(file, ctx)) 3651 return false; 3652 ctx->pos = 1; 3653 } 3654 if (ctx->pos == 1) { 3655 if (!dir_emit_dotdot(file, ctx)) 3656 return false; 3657 ctx->pos = 2; 3658 } 3659 return true; 3660 } 3661 static inline bool dir_relax(struct inode *inode) 3662 { 3663 inode_unlock(inode); 3664 inode_lock(inode); 3665 return !IS_DEADDIR(inode); 3666 } 3667 3668 static inline bool dir_relax_shared(struct inode *inode) 3669 { 3670 inode_unlock_shared(inode); 3671 inode_lock_shared(inode); 3672 return !IS_DEADDIR(inode); 3673 } 3674 3675 extern bool path_noexec(const struct path *path); 3676 extern void inode_nohighmem(struct inode *inode); 3677 3678 /* mm/fadvise.c */ 3679 extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len, 3680 int advice); 3681 extern int generic_fadvise(struct file *file, loff_t offset, loff_t len, 3682 int advice); 3683 3684 /* 3685 * Flush file data before changing attributes. Caller must hold any locks 3686 * required to prevent further writes to this file until we're done setting 3687 * flags. 3688 */ 3689 static inline int inode_drain_writes(struct inode *inode) 3690 { 3691 inode_dio_wait(inode); 3692 return filemap_write_and_wait(inode->i_mapping); 3693 } 3694 3695 #endif /* _LINUX_FS_H */ 3696