linux/include/linux/fs.h
<<
>>
Prefs
   1#ifndef _LINUX_FS_H
   2#define _LINUX_FS_H
   3
   4#include <linux/linkage.h>
   5#include <linux/wait_bit.h>
   6#include <linux/kdev_t.h>
   7#include <linux/dcache.h>
   8#include <linux/path.h>
   9#include <linux/stat.h>
  10#include <linux/cache.h>
  11#include <linux/list.h>
  12#include <linux/list_lru.h>
  13#include <linux/llist.h>
  14#include <linux/radix-tree.h>
  15#include <linux/rbtree.h>
  16#include <linux/init.h>
  17#include <linux/pid.h>
  18#include <linux/bug.h>
  19#include <linux/mutex.h>
  20#include <linux/rwsem.h>
  21#include <linux/mm_types.h>
  22#include <linux/capability.h>
  23#include <linux/semaphore.h>
  24#include <linux/fcntl.h>
  25#include <linux/fiemap.h>
  26#include <linux/rculist_bl.h>
  27#include <linux/atomic.h>
  28#include <linux/shrinker.h>
  29#include <linux/migrate_mode.h>
  30#include <linux/uidgid.h>
  31#include <linux/lockdep.h>
  32#include <linux/percpu-rwsem.h>
  33#include <linux/workqueue.h>
  34#include <linux/delayed_call.h>
  35#include <linux/uuid.h>
  36#include <linux/errseq.h>
  37
  38#include <asm/byteorder.h>
  39#include <uapi/linux/fs.h>
  40
  41struct backing_dev_info;
  42struct bdi_writeback;
  43struct bio;
  44struct export_operations;
  45struct hd_geometry;
  46struct iovec;
  47struct kiocb;
  48struct kobject;
  49struct pipe_inode_info;
  50struct poll_table_struct;
  51struct kstatfs;
  52struct vm_area_struct;
  53struct vfsmount;
  54struct cred;
  55struct swap_info_struct;
  56struct seq_file;
  57struct workqueue_struct;
  58struct iov_iter;
  59struct fscrypt_info;
  60struct fscrypt_operations;
  61
  62extern void __init inode_init(void);
  63extern void __init inode_init_early(void);
  64extern void __init files_init(void);
  65extern void __init files_maxfiles_init(void);
  66
  67extern struct files_stat_struct files_stat;
  68extern unsigned long get_max_files(void);
  69extern unsigned int sysctl_nr_open;
  70extern struct inodes_stat_t inodes_stat;
  71extern int leases_enable, lease_break_time;
  72extern int sysctl_protected_symlinks;
  73extern int sysctl_protected_hardlinks;
  74
  75struct buffer_head;
  76typedef int (get_block_t)(struct inode *inode, sector_t iblock,
  77                        struct buffer_head *bh_result, int create);
  78typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
  79                        ssize_t bytes, void *private);
  80
  81#define MAY_EXEC                0x00000001
  82#define MAY_WRITE               0x00000002
  83#define MAY_READ                0x00000004
  84#define MAY_APPEND              0x00000008
  85#define MAY_ACCESS              0x00000010
  86#define MAY_OPEN                0x00000020
  87#define MAY_CHDIR               0x00000040
  88/* called from RCU mode, don't block */
  89#define MAY_NOT_BLOCK           0x00000080
  90
  91/*
  92 * flags in file.f_mode.  Note that FMODE_READ and FMODE_WRITE must correspond
  93 * to O_WRONLY and O_RDWR via the strange trick in __dentry_open()
  94 */
  95
  96/* file is open for reading */
  97#define FMODE_READ              ((__force fmode_t)0x1)
  98/* file is open for writing */
  99#define FMODE_WRITE             ((__force fmode_t)0x2)
 100/* file is seekable */
 101#define FMODE_LSEEK             ((__force fmode_t)0x4)
 102/* file can be accessed using pread */
 103#define FMODE_PREAD             ((__force fmode_t)0x8)
 104/* file can be accessed using pwrite */
 105#define FMODE_PWRITE            ((__force fmode_t)0x10)
 106/* File is opened for execution with sys_execve / sys_uselib */
 107#define FMODE_EXEC              ((__force fmode_t)0x20)
 108/* File is opened with O_NDELAY (only set for block devices) */
 109#define FMODE_NDELAY            ((__force fmode_t)0x40)
 110/* File is opened with O_EXCL (only set for block devices) */
 111#define FMODE_EXCL              ((__force fmode_t)0x80)
 112/* File is opened using open(.., 3, ..) and is writeable only for ioctls
 113   (specialy hack for floppy.c) */
 114#define FMODE_WRITE_IOCTL       ((__force fmode_t)0x100)
 115/* 32bit hashes as llseek() offset (for directories) */
 116#define FMODE_32BITHASH         ((__force fmode_t)0x200)
 117/* 64bit hashes as llseek() offset (for directories) */
 118#define FMODE_64BITHASH         ((__force fmode_t)0x400)
 119
 120/*
 121 * Don't update ctime and mtime.
 122 *
 123 * Currently a special hack for the XFS open_by_handle ioctl, but we'll
 124 * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon.
 125 */
 126#define FMODE_NOCMTIME          ((__force fmode_t)0x800)
 127
 128/* Expect random access pattern */
 129#define FMODE_RANDOM            ((__force fmode_t)0x1000)
 130
 131/* File is huge (eg. /dev/kmem): treat loff_t as unsigned */
 132#define FMODE_UNSIGNED_OFFSET   ((__force fmode_t)0x2000)
 133
 134/* File is opened with O_PATH; almost nothing can be done with it */
 135#define FMODE_PATH              ((__force fmode_t)0x4000)
 136
 137/* File needs atomic accesses to f_pos */
 138#define FMODE_ATOMIC_POS        ((__force fmode_t)0x8000)
 139/* Write access to underlying fs */
 140#define FMODE_WRITER            ((__force fmode_t)0x10000)
 141/* Has read method(s) */
 142#define FMODE_CAN_READ          ((__force fmode_t)0x20000)
 143/* Has write method(s) */
 144#define FMODE_CAN_WRITE         ((__force fmode_t)0x40000)
 145
 146/* File was opened by fanotify and shouldn't generate fanotify events */
 147#define FMODE_NONOTIFY          ((__force fmode_t)0x4000000)
 148
 149/* File is capable of returning -EAGAIN if AIO will block */
 150#define FMODE_AIO_NOWAIT        ((__force fmode_t)0x8000000)
 151
 152/*
 153 * Flag for rw_copy_check_uvector and compat_rw_copy_check_uvector
 154 * that indicates that they should check the contents of the iovec are
 155 * valid, but not check the memory that the iovec elements
 156 * points too.
 157 */
 158#define CHECK_IOVEC_ONLY -1
 159
 160/*
 161 * Attribute flags.  These should be or-ed together to figure out what
 162 * has been changed!
 163 */
 164#define ATTR_MODE       (1 << 0)
 165#define ATTR_UID        (1 << 1)
 166#define ATTR_GID        (1 << 2)
 167#define ATTR_SIZE       (1 << 3)
 168#define ATTR_ATIME      (1 << 4)
 169#define ATTR_MTIME      (1 << 5)
 170#define ATTR_CTIME      (1 << 6)
 171#define ATTR_ATIME_SET  (1 << 7)
 172#define ATTR_MTIME_SET  (1 << 8)
 173#define ATTR_FORCE      (1 << 9) /* Not a change, but a change it */
 174#define ATTR_ATTR_FLAG  (1 << 10)
 175#define ATTR_KILL_SUID  (1 << 11)
 176#define ATTR_KILL_SGID  (1 << 12)
 177#define ATTR_FILE       (1 << 13)
 178#define ATTR_KILL_PRIV  (1 << 14)
 179#define ATTR_OPEN       (1 << 15) /* Truncating from open(O_TRUNC) */
 180#define ATTR_TIMES_SET  (1 << 16)
 181#define ATTR_TOUCH      (1 << 17)
 182
 183/*
 184 * Whiteout is represented by a char device.  The following constants define the
 185 * mode and device number to use.
 186 */
 187#define WHITEOUT_MODE 0
 188#define WHITEOUT_DEV 0
 189
 190/*
 191 * This is the Inode Attributes structure, used for notify_change().  It
 192 * uses the above definitions as flags, to know which values have changed.
 193 * Also, in this manner, a Filesystem can look at only the values it cares
 194 * about.  Basically, these are the attributes that the VFS layer can
 195 * request to change from the FS layer.
 196 *
 197 * Derek Atkins <warlord@MIT.EDU> 94-10-20
 198 */
 199struct iattr {
 200        unsigned int    ia_valid;
 201        umode_t         ia_mode;
 202        kuid_t          ia_uid;
 203        kgid_t          ia_gid;
 204        loff_t          ia_size;
 205        struct timespec ia_atime;
 206        struct timespec ia_mtime;
 207        struct timespec ia_ctime;
 208
 209        /*
 210         * Not an attribute, but an auxiliary info for filesystems wanting to
 211         * implement an ftruncate() like method.  NOTE: filesystem should
 212         * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL).
 213         */
 214        struct file     *ia_file;
 215};
 216
 217/*
 218 * Includes for diskquotas.
 219 */
 220#include <linux/quota.h>
 221
 222/*
 223 * Maximum number of layers of fs stack.  Needs to be limited to
 224 * prevent kernel stack overflow
 225 */
 226#define FILESYSTEM_MAX_STACK_DEPTH 2
 227
 228/** 
 229 * enum positive_aop_returns - aop return codes with specific semantics
 230 *
 231 * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has
 232 *                          completed, that the page is still locked, and
 233 *                          should be considered active.  The VM uses this hint
 234 *                          to return the page to the active list -- it won't
 235 *                          be a candidate for writeback again in the near
 236 *                          future.  Other callers must be careful to unlock
 237 *                          the page if they get this return.  Returned by
 238 *                          writepage(); 
 239 *
 240 * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has
 241 *                      unlocked it and the page might have been truncated.
 242 *                      The caller should back up to acquiring a new page and
 243 *                      trying again.  The aop will be taking reasonable
 244 *                      precautions not to livelock.  If the caller held a page
 245 *                      reference, it should drop it before retrying.  Returned
 246 *                      by readpage().
 247 *
 248 * address_space_operation functions return these large constants to indicate
 249 * special semantics to the caller.  These are much larger than the bytes in a
 250 * page to allow for functions that return the number of bytes operated on in a
 251 * given page.
 252 */
 253
 254enum positive_aop_returns {
 255        AOP_WRITEPAGE_ACTIVATE  = 0x80000,
 256        AOP_TRUNCATED_PAGE      = 0x80001,
 257};
 258
 259#define AOP_FLAG_CONT_EXPAND            0x0001 /* called from cont_expand */
 260#define AOP_FLAG_NOFS                   0x0002 /* used by filesystem to direct
 261                                                * helper code (eg buffer layer)
 262                                                * to clear GFP_FS from alloc */
 263
 264/*
 265 * oh the beauties of C type declarations.
 266 */
 267struct page;
 268struct address_space;
 269struct writeback_control;
 270
 271/*
 272 * Write life time hint values.
 273 */
 274enum rw_hint {
 275        WRITE_LIFE_NOT_SET      = 0,
 276        WRITE_LIFE_NONE         = RWH_WRITE_LIFE_NONE,
 277        WRITE_LIFE_SHORT        = RWH_WRITE_LIFE_SHORT,
 278        WRITE_LIFE_MEDIUM       = RWH_WRITE_LIFE_MEDIUM,
 279        WRITE_LIFE_LONG         = RWH_WRITE_LIFE_LONG,
 280        WRITE_LIFE_EXTREME      = RWH_WRITE_LIFE_EXTREME,
 281};
 282
 283#define IOCB_EVENTFD            (1 << 0)
 284#define IOCB_APPEND             (1 << 1)
 285#define IOCB_DIRECT             (1 << 2)
 286#define IOCB_HIPRI              (1 << 3)
 287#define IOCB_DSYNC              (1 << 4)
 288#define IOCB_SYNC               (1 << 5)
 289#define IOCB_WRITE              (1 << 6)
 290#define IOCB_NOWAIT             (1 << 7)
 291
 292struct kiocb {
 293        struct file             *ki_filp;
 294        loff_t                  ki_pos;
 295        void (*ki_complete)(struct kiocb *iocb, long ret, long ret2);
 296        void                    *private;
 297        int                     ki_flags;
 298        enum rw_hint            ki_hint;
 299} __randomize_layout;
 300
 301static inline bool is_sync_kiocb(struct kiocb *kiocb)
 302{
 303        return kiocb->ki_complete == NULL;
 304}
 305
 306/*
 307 * "descriptor" for what we're up to with a read.
 308 * This allows us to use the same read code yet
 309 * have multiple different users of the data that
 310 * we read from a file.
 311 *
 312 * The simplest case just copies the data to user
 313 * mode.
 314 */
 315typedef struct {
 316        size_t written;
 317        size_t count;
 318        union {
 319                char __user *buf;
 320                void *data;
 321        } arg;
 322        int error;
 323} read_descriptor_t;
 324
 325typedef int (*read_actor_t)(read_descriptor_t *, struct page *,
 326                unsigned long, unsigned long);
 327
 328struct address_space_operations {
 329        int (*writepage)(struct page *page, struct writeback_control *wbc);
 330        int (*readpage)(struct file *, struct page *);
 331
 332        /* Write back some dirty pages from this mapping. */
 333        int (*writepages)(struct address_space *, struct writeback_control *);
 334
 335        /* Set a page dirty.  Return true if this dirtied it */
 336        int (*set_page_dirty)(struct page *page);
 337
 338        int (*readpages)(struct file *filp, struct address_space *mapping,
 339                        struct list_head *pages, unsigned nr_pages);
 340
 341        int (*write_begin)(struct file *, struct address_space *mapping,
 342                                loff_t pos, unsigned len, unsigned flags,
 343                                struct page **pagep, void **fsdata);
 344        int (*write_end)(struct file *, struct address_space *mapping,
 345                                loff_t pos, unsigned len, unsigned copied,
 346                                struct page *page, void *fsdata);
 347
 348        /* Unfortunately this kludge is needed for FIBMAP. Don't use it */
 349        sector_t (*bmap)(struct address_space *, sector_t);
 350        void (*invalidatepage) (struct page *, unsigned int, unsigned int);
 351        int (*releasepage) (struct page *, gfp_t);
 352        void (*freepage)(struct page *);
 353        ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter);
 354        /*
 355         * migrate the contents of a page to the specified target. If
 356         * migrate_mode is MIGRATE_ASYNC, it must not block.
 357         */
 358        int (*migratepage) (struct address_space *,
 359                        struct page *, struct page *, enum migrate_mode);
 360        bool (*isolate_page)(struct page *, isolate_mode_t);
 361        void (*putback_page)(struct page *);
 362        int (*launder_page) (struct page *);
 363        int (*is_partially_uptodate) (struct page *, unsigned long,
 364                                        unsigned long);
 365        void (*is_dirty_writeback) (struct page *, bool *, bool *);
 366        int (*error_remove_page)(struct address_space *, struct page *);
 367
 368        /* swapfile support */
 369        int (*swap_activate)(struct swap_info_struct *sis, struct file *file,
 370                                sector_t *span);
 371        void (*swap_deactivate)(struct file *file);
 372};
 373
 374extern const struct address_space_operations empty_aops;
 375
 376/*
 377 * pagecache_write_begin/pagecache_write_end must be used by general code
 378 * to write into the pagecache.
 379 */
 380int pagecache_write_begin(struct file *, struct address_space *mapping,
 381                                loff_t pos, unsigned len, unsigned flags,
 382                                struct page **pagep, void **fsdata);
 383
 384int pagecache_write_end(struct file *, struct address_space *mapping,
 385                                loff_t pos, unsigned len, unsigned copied,
 386                                struct page *page, void *fsdata);
 387
 388struct address_space {
 389        struct inode            *host;          /* owner: inode, block_device */
 390        struct radix_tree_root  page_tree;      /* radix tree of all pages */
 391        spinlock_t              tree_lock;      /* and lock protecting it */
 392        atomic_t                i_mmap_writable;/* count VM_SHARED mappings */
 393        struct rb_root          i_mmap;         /* tree of private and shared mappings */
 394        struct rw_semaphore     i_mmap_rwsem;   /* protect tree, count, list */
 395        /* Protected by tree_lock together with the radix tree */
 396        unsigned long           nrpages;        /* number of total pages */
 397        /* number of shadow or DAX exceptional entries */
 398        unsigned long           nrexceptional;
 399        pgoff_t                 writeback_index;/* writeback starts here */
 400        const struct address_space_operations *a_ops;   /* methods */
 401        unsigned long           flags;          /* error bits */
 402        spinlock_t              private_lock;   /* for use by the address_space */
 403        gfp_t                   gfp_mask;       /* implicit gfp mask for allocations */
 404        struct list_head        private_list;   /* ditto */
 405        void                    *private_data;  /* ditto */
 406        errseq_t                wb_err;
 407} __attribute__((aligned(sizeof(long)))) __randomize_layout;
 408        /*
 409         * On most architectures that alignment is already the case; but
 410         * must be enforced here for CRIS, to let the least significant bit
 411         * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON.
 412         */
 413struct request_queue;
 414
 415struct block_device {
 416        dev_t                   bd_dev;  /* not a kdev_t - it's a search key */
 417        int                     bd_openers;
 418        struct inode *          bd_inode;       /* will die */
 419        struct super_block *    bd_super;
 420        struct mutex            bd_mutex;       /* open/close mutex */
 421        void *                  bd_claiming;
 422        void *                  bd_holder;
 423        int                     bd_holders;
 424        bool                    bd_write_holder;
 425#ifdef CONFIG_SYSFS
 426        struct list_head        bd_holder_disks;
 427#endif
 428        struct block_device *   bd_contains;
 429        unsigned                bd_block_size;
 430        struct hd_struct *      bd_part;
 431        /* number of times partitions within this device have been opened. */
 432        unsigned                bd_part_count;
 433        int                     bd_invalidated;
 434        struct gendisk *        bd_disk;
 435        struct request_queue *  bd_queue;
 436        struct backing_dev_info *bd_bdi;
 437        struct list_head        bd_list;
 438        /*
 439         * Private data.  You must have bd_claim'ed the block_device
 440         * to use this.  NOTE:  bd_claim allows an owner to claim
 441         * the same device multiple times, the owner must take special
 442         * care to not mess up bd_private for that case.
 443         */
 444        unsigned long           bd_private;
 445
 446        /* The counter of freeze processes */
 447        int                     bd_fsfreeze_count;
 448        /* Mutex for freeze */
 449        struct mutex            bd_fsfreeze_mutex;
 450} __randomize_layout;
 451
 452/*
 453 * Radix-tree tags, for tagging dirty and writeback pages within the pagecache
 454 * radix trees
 455 */
 456#define PAGECACHE_TAG_DIRTY     0
 457#define PAGECACHE_TAG_WRITEBACK 1
 458#define PAGECACHE_TAG_TOWRITE   2
 459
 460int mapping_tagged(struct address_space *mapping, int tag);
 461
 462static inline void i_mmap_lock_write(struct address_space *mapping)
 463{
 464        down_write(&mapping->i_mmap_rwsem);
 465}
 466
 467static inline void i_mmap_unlock_write(struct address_space *mapping)
 468{
 469        up_write(&mapping->i_mmap_rwsem);
 470}
 471
 472static inline void i_mmap_lock_read(struct address_space *mapping)
 473{
 474        down_read(&mapping->i_mmap_rwsem);
 475}
 476
 477static inline void i_mmap_unlock_read(struct address_space *mapping)
 478{
 479        up_read(&mapping->i_mmap_rwsem);
 480}
 481
 482/*
 483 * Might pages of this file be mapped into userspace?
 484 */
 485static inline int mapping_mapped(struct address_space *mapping)
 486{
 487        return  !RB_EMPTY_ROOT(&mapping->i_mmap);
 488}
 489
 490/*
 491 * Might pages of this file have been modified in userspace?
 492 * Note that i_mmap_writable counts all VM_SHARED vmas: do_mmap_pgoff
 493 * marks vma as VM_SHARED if it is shared, and the file was opened for
 494 * writing i.e. vma may be mprotected writable even if now readonly.
 495 *
 496 * If i_mmap_writable is negative, no new writable mappings are allowed. You
 497 * can only deny writable mappings, if none exists right now.
 498 */
 499static inline int mapping_writably_mapped(struct address_space *mapping)
 500{
 501        return atomic_read(&mapping->i_mmap_writable) > 0;
 502}
 503
 504static inline int mapping_map_writable(struct address_space *mapping)
 505{
 506        return atomic_inc_unless_negative(&mapping->i_mmap_writable) ?
 507                0 : -EPERM;
 508}
 509
 510static inline void mapping_unmap_writable(struct address_space *mapping)
 511{
 512        atomic_dec(&mapping->i_mmap_writable);
 513}
 514
 515static inline int mapping_deny_writable(struct address_space *mapping)
 516{
 517        return atomic_dec_unless_positive(&mapping->i_mmap_writable) ?
 518                0 : -EBUSY;
 519}
 520
 521static inline void mapping_allow_writable(struct address_space *mapping)
 522{
 523        atomic_inc(&mapping->i_mmap_writable);
 524}
 525
 526/*
 527 * Use sequence counter to get consistent i_size on 32-bit processors.
 528 */
 529#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
 530#include <linux/seqlock.h>
 531#define __NEED_I_SIZE_ORDERED
 532#define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount)
 533#else
 534#define i_size_ordered_init(inode) do { } while (0)
 535#endif
 536
 537struct posix_acl;
 538#define ACL_NOT_CACHED ((void *)(-1))
 539#define ACL_DONT_CACHE ((void *)(-3))
 540
 541static inline struct posix_acl *
 542uncached_acl_sentinel(struct task_struct *task)
 543{
 544        return (void *)task + 1;
 545}
 546
 547static inline bool
 548is_uncached_acl(struct posix_acl *acl)
 549{
 550        return (long)acl & 1;
 551}
 552
 553#define IOP_FASTPERM    0x0001
 554#define IOP_LOOKUP      0x0002
 555#define IOP_NOFOLLOW    0x0004
 556#define IOP_XATTR       0x0008
 557#define IOP_DEFAULT_READLINK    0x0010
 558
 559struct fsnotify_mark_connector;
 560
 561/*
 562 * Keep mostly read-only and often accessed (especially for
 563 * the RCU path lookup and 'stat' data) fields at the beginning
 564 * of the 'struct inode'
 565 */
 566struct inode {
 567        umode_t                 i_mode;
 568        unsigned short          i_opflags;
 569        kuid_t                  i_uid;
 570        kgid_t                  i_gid;
 571        unsigned int            i_flags;
 572
 573#ifdef CONFIG_FS_POSIX_ACL
 574        struct posix_acl        *i_acl;
 575        struct posix_acl        *i_default_acl;
 576#endif
 577
 578        const struct inode_operations   *i_op;
 579        struct super_block      *i_sb;
 580        struct address_space    *i_mapping;
 581
 582#ifdef CONFIG_SECURITY
 583        void                    *i_security;
 584#endif
 585
 586        /* Stat data, not accessed from path walking */
 587        unsigned long           i_ino;
 588        /*
 589         * Filesystems may only read i_nlink directly.  They shall use the
 590         * following functions for modification:
 591         *
 592         *    (set|clear|inc|drop)_nlink
 593         *    inode_(inc|dec)_link_count
 594         */
 595        union {
 596                const unsigned int i_nlink;
 597                unsigned int __i_nlink;
 598        };
 599        dev_t                   i_rdev;
 600        loff_t                  i_size;
 601        struct timespec         i_atime;
 602        struct timespec         i_mtime;
 603        struct timespec         i_ctime;
 604        spinlock_t              i_lock; /* i_blocks, i_bytes, maybe i_size */
 605        unsigned short          i_bytes;
 606        unsigned int            i_blkbits;
 607        enum rw_hint            i_write_hint;
 608        blkcnt_t                i_blocks;
 609
 610#ifdef __NEED_I_SIZE_ORDERED
 611        seqcount_t              i_size_seqcount;
 612#endif
 613
 614        /* Misc */
 615        unsigned long           i_state;
 616        struct rw_semaphore     i_rwsem;
 617
 618        unsigned long           dirtied_when;   /* jiffies of first dirtying */
 619        unsigned long           dirtied_time_when;
 620
 621        struct hlist_node       i_hash;
 622        struct list_head        i_io_list;      /* backing dev IO list */
 623#ifdef CONFIG_CGROUP_WRITEBACK
 624        struct bdi_writeback    *i_wb;          /* the associated cgroup wb */
 625
 626        /* foreign inode detection, see wbc_detach_inode() */
 627        int                     i_wb_frn_winner;
 628        u16                     i_wb_frn_avg_time;
 629        u16                     i_wb_frn_history;
 630#endif
 631        struct list_head        i_lru;          /* inode LRU list */
 632        struct list_head        i_sb_list;
 633        struct list_head        i_wb_list;      /* backing dev writeback list */
 634        union {
 635                struct hlist_head       i_dentry;
 636                struct rcu_head         i_rcu;
 637        };
 638        u64                     i_version;
 639        atomic_t                i_count;
 640        atomic_t                i_dio_count;
 641        atomic_t                i_writecount;
 642#ifdef CONFIG_IMA
 643        atomic_t                i_readcount; /* struct files open RO */
 644#endif
 645        const struct file_operations    *i_fop; /* former ->i_op->default_file_ops */
 646        struct file_lock_context        *i_flctx;
 647        struct address_space    i_data;
 648        struct list_head        i_devices;
 649        union {
 650                struct pipe_inode_info  *i_pipe;
 651                struct block_device     *i_bdev;
 652                struct cdev             *i_cdev;
 653                char                    *i_link;
 654                unsigned                i_dir_seq;
 655        };
 656
 657        __u32                   i_generation;
 658
 659#ifdef CONFIG_FSNOTIFY
 660        __u32                   i_fsnotify_mask; /* all events this inode cares about */
 661        struct fsnotify_mark_connector __rcu    *i_fsnotify_marks;
 662#endif
 663
 664#if IS_ENABLED(CONFIG_FS_ENCRYPTION)
 665        struct fscrypt_info     *i_crypt_info;
 666#endif
 667
 668        void                    *i_private; /* fs or device private pointer */
 669} __randomize_layout;
 670
 671static inline unsigned int i_blocksize(const struct inode *node)
 672{
 673        return (1 << node->i_blkbits);
 674}
 675
 676static inline int inode_unhashed(struct inode *inode)
 677{
 678        return hlist_unhashed(&inode->i_hash);
 679}
 680
 681/*
 682 * inode->i_mutex nesting subclasses for the lock validator:
 683 *
 684 * 0: the object of the current VFS operation
 685 * 1: parent
 686 * 2: child/target
 687 * 3: xattr
 688 * 4: second non-directory
 689 * 5: second parent (when locking independent directories in rename)
 690 *
 691 * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two
 692 * non-directories at once.
 693 *
 694 * The locking order between these classes is
 695 * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory
 696 */
 697enum inode_i_mutex_lock_class
 698{
 699        I_MUTEX_NORMAL,
 700        I_MUTEX_PARENT,
 701        I_MUTEX_CHILD,
 702        I_MUTEX_XATTR,
 703        I_MUTEX_NONDIR2,
 704        I_MUTEX_PARENT2,
 705};
 706
 707static inline void inode_lock(struct inode *inode)
 708{
 709        down_write(&inode->i_rwsem);
 710}
 711
 712static inline void inode_unlock(struct inode *inode)
 713{
 714        up_write(&inode->i_rwsem);
 715}
 716
 717static inline void inode_lock_shared(struct inode *inode)
 718{
 719        down_read(&inode->i_rwsem);
 720}
 721
 722static inline void inode_unlock_shared(struct inode *inode)
 723{
 724        up_read(&inode->i_rwsem);
 725}
 726
 727static inline int inode_trylock(struct inode *inode)
 728{
 729        return down_write_trylock(&inode->i_rwsem);
 730}
 731
 732static inline int inode_trylock_shared(struct inode *inode)
 733{
 734        return down_read_trylock(&inode->i_rwsem);
 735}
 736
 737static inline int inode_is_locked(struct inode *inode)
 738{
 739        return rwsem_is_locked(&inode->i_rwsem);
 740}
 741
 742static inline void inode_lock_nested(struct inode *inode, unsigned subclass)
 743{
 744        down_write_nested(&inode->i_rwsem, subclass);
 745}
 746
 747void lock_two_nondirectories(struct inode *, struct inode*);
 748void unlock_two_nondirectories(struct inode *, struct inode*);
 749
 750/*
 751 * NOTE: in a 32bit arch with a preemptable kernel and
 752 * an UP compile the i_size_read/write must be atomic
 753 * with respect to the local cpu (unlike with preempt disabled),
 754 * but they don't need to be atomic with respect to other cpus like in
 755 * true SMP (so they need either to either locally disable irq around
 756 * the read or for example on x86 they can be still implemented as a
 757 * cmpxchg8b without the need of the lock prefix). For SMP compiles
 758 * and 64bit archs it makes no difference if preempt is enabled or not.
 759 */
 760static inline loff_t i_size_read(const struct inode *inode)
 761{
 762#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
 763        loff_t i_size;
 764        unsigned int seq;
 765
 766        do {
 767                seq = read_seqcount_begin(&inode->i_size_seqcount);
 768                i_size = inode->i_size;
 769        } while (read_seqcount_retry(&inode->i_size_seqcount, seq));
 770        return i_size;
 771#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
 772        loff_t i_size;
 773
 774        preempt_disable();
 775        i_size = inode->i_size;
 776        preempt_enable();
 777        return i_size;
 778#else
 779        return inode->i_size;
 780#endif
 781}
 782
 783/*
 784 * NOTE: unlike i_size_read(), i_size_write() does need locking around it
 785 * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount
 786 * can be lost, resulting in subsequent i_size_read() calls spinning forever.
 787 */
 788static inline void i_size_write(struct inode *inode, loff_t i_size)
 789{
 790#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
 791        preempt_disable();
 792        write_seqcount_begin(&inode->i_size_seqcount);
 793        inode->i_size = i_size;
 794        write_seqcount_end(&inode->i_size_seqcount);
 795        preempt_enable();
 796#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPT)
 797        preempt_disable();
 798        inode->i_size = i_size;
 799        preempt_enable();
 800#else
 801        inode->i_size = i_size;
 802#endif
 803}
 804
 805static inline unsigned iminor(const struct inode *inode)
 806{
 807        return MINOR(inode->i_rdev);
 808}
 809
 810static inline unsigned imajor(const struct inode *inode)
 811{
 812        return MAJOR(inode->i_rdev);
 813}
 814
 815extern struct block_device *I_BDEV(struct inode *inode);
 816
 817struct fown_struct {
 818        rwlock_t lock;          /* protects pid, uid, euid fields */
 819        struct pid *pid;        /* pid or -pgrp where SIGIO should be sent */
 820        enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */
 821        kuid_t uid, euid;       /* uid/euid of process setting the owner */
 822        int signum;             /* posix.1b rt signal to be delivered on IO */
 823};
 824
 825/*
 826 * Track a single file's readahead state
 827 */
 828struct file_ra_state {
 829        pgoff_t start;                  /* where readahead started */
 830        unsigned int size;              /* # of readahead pages */
 831        unsigned int async_size;        /* do asynchronous readahead when
 832                                           there are only # of pages ahead */
 833
 834        unsigned int ra_pages;          /* Maximum readahead window */
 835        unsigned int mmap_miss;         /* Cache miss stat for mmap accesses */
 836        loff_t prev_pos;                /* Cache last read() position */
 837};
 838
 839/*
 840 * Check if @index falls in the readahead windows.
 841 */
 842static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index)
 843{
 844        return (index >= ra->start &&
 845                index <  ra->start + ra->size);
 846}
 847
 848struct file {
 849        union {
 850                struct llist_node       fu_llist;
 851                struct rcu_head         fu_rcuhead;
 852        } f_u;
 853        struct path             f_path;
 854        struct inode            *f_inode;       /* cached value */
 855        const struct file_operations    *f_op;
 856
 857        /*
 858         * Protects f_ep_links, f_flags.
 859         * Must not be taken from IRQ context.
 860         */
 861        spinlock_t              f_lock;
 862        enum rw_hint            f_write_hint;
 863        atomic_long_t           f_count;
 864        unsigned int            f_flags;
 865        fmode_t                 f_mode;
 866        struct mutex            f_pos_lock;
 867        loff_t                  f_pos;
 868        struct fown_struct      f_owner;
 869        const struct cred       *f_cred;
 870        struct file_ra_state    f_ra;
 871
 872        u64                     f_version;
 873#ifdef CONFIG_SECURITY
 874        void                    *f_security;
 875#endif
 876        /* needed for tty driver, and maybe others */
 877        void                    *private_data;
 878
 879#ifdef CONFIG_EPOLL
 880        /* Used by fs/eventpoll.c to link all the hooks to this file */
 881        struct list_head        f_ep_links;
 882        struct list_head        f_tfile_llink;
 883#endif /* #ifdef CONFIG_EPOLL */
 884        struct address_space    *f_mapping;
 885        errseq_t                f_wb_err;
 886} __randomize_layout
 887  __attribute__((aligned(4)));  /* lest something weird decides that 2 is OK */
 888
 889struct file_handle {
 890        __u32 handle_bytes;
 891        int handle_type;
 892        /* file identifier */
 893        unsigned char f_handle[0];
 894};
 895
 896static inline struct file *get_file(struct file *f)
 897{
 898        atomic_long_inc(&f->f_count);
 899        return f;
 900}
 901#define get_file_rcu(x) atomic_long_inc_not_zero(&(x)->f_count)
 902#define fput_atomic(x)  atomic_long_add_unless(&(x)->f_count, -1, 1)
 903#define file_count(x)   atomic_long_read(&(x)->f_count)
 904
 905#define MAX_NON_LFS     ((1UL<<31) - 1)
 906
 907/* Page cache limit. The filesystems should put that into their s_maxbytes 
 908   limits, otherwise bad things can happen in VM. */ 
 909#if BITS_PER_LONG==32
 910#define MAX_LFS_FILESIZE        ((loff_t)ULONG_MAX << PAGE_SHIFT)
 911#elif BITS_PER_LONG==64
 912#define MAX_LFS_FILESIZE        ((loff_t)LLONG_MAX)
 913#endif
 914
 915#define FL_POSIX        1
 916#define FL_FLOCK        2
 917#define FL_DELEG        4       /* NFSv4 delegation */
 918#define FL_ACCESS       8       /* not trying to lock, just looking */
 919#define FL_EXISTS       16      /* when unlocking, test for existence */
 920#define FL_LEASE        32      /* lease held on this file */
 921#define FL_CLOSE        64      /* unlock on close */
 922#define FL_SLEEP        128     /* A blocking lock */
 923#define FL_DOWNGRADE_PENDING    256 /* Lease is being downgraded */
 924#define FL_UNLOCK_PENDING       512 /* Lease is being broken */
 925#define FL_OFDLCK       1024    /* lock is "owned" by struct file */
 926#define FL_LAYOUT       2048    /* outstanding pNFS layout */
 927
 928#define FL_CLOSE_POSIX (FL_POSIX | FL_CLOSE)
 929
 930/*
 931 * Special return value from posix_lock_file() and vfs_lock_file() for
 932 * asynchronous locking.
 933 */
 934#define FILE_LOCK_DEFERRED 1
 935
 936/* legacy typedef, should eventually be removed */
 937typedef void *fl_owner_t;
 938
 939struct file_lock;
 940
 941struct file_lock_operations {
 942        void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
 943        void (*fl_release_private)(struct file_lock *);
 944};
 945
 946struct lock_manager_operations {
 947        int (*lm_compare_owner)(struct file_lock *, struct file_lock *);
 948        unsigned long (*lm_owner_key)(struct file_lock *);
 949        fl_owner_t (*lm_get_owner)(fl_owner_t);
 950        void (*lm_put_owner)(fl_owner_t);
 951        void (*lm_notify)(struct file_lock *);  /* unblock callback */
 952        int (*lm_grant)(struct file_lock *, int);
 953        bool (*lm_break)(struct file_lock *);
 954        int (*lm_change)(struct file_lock *, int, struct list_head *);
 955        void (*lm_setup)(struct file_lock *, void **);
 956};
 957
 958struct lock_manager {
 959        struct list_head list;
 960        /*
 961         * NFSv4 and up also want opens blocked during the grace period;
 962         * NLM doesn't care:
 963         */
 964        bool block_opens;
 965};
 966
 967struct net;
 968void locks_start_grace(struct net *, struct lock_manager *);
 969void locks_end_grace(struct lock_manager *);
 970int locks_in_grace(struct net *);
 971int opens_in_grace(struct net *);
 972
 973/* that will die - we need it for nfs_lock_info */
 974#include <linux/nfs_fs_i.h>
 975
 976/*
 977 * struct file_lock represents a generic "file lock". It's used to represent
 978 * POSIX byte range locks, BSD (flock) locks, and leases. It's important to
 979 * note that the same struct is used to represent both a request for a lock and
 980 * the lock itself, but the same object is never used for both.
 981 *
 982 * FIXME: should we create a separate "struct lock_request" to help distinguish
 983 * these two uses?
 984 *
 985 * The varous i_flctx lists are ordered by:
 986 *
 987 * 1) lock owner
 988 * 2) lock range start
 989 * 3) lock range end
 990 *
 991 * Obviously, the last two criteria only matter for POSIX locks.
 992 */
 993struct file_lock {
 994        struct file_lock *fl_next;      /* singly linked list for this inode  */
 995        struct list_head fl_list;       /* link into file_lock_context */
 996        struct hlist_node fl_link;      /* node in global lists */
 997        struct list_head fl_block;      /* circular list of blocked processes */
 998        fl_owner_t fl_owner;
 999        unsigned int fl_flags;
1000        unsigned char fl_type;
1001        unsigned int fl_pid;
1002        int fl_link_cpu;                /* what cpu's list is this on? */
1003        struct pid *fl_nspid;
1004        wait_queue_head_t fl_wait;
1005        struct file *fl_file;
1006        loff_t fl_start;
1007        loff_t fl_end;
1008
1009        struct fasync_struct *  fl_fasync; /* for lease break notifications */
1010        /* for lease breaks: */
1011        unsigned long fl_break_time;
1012        unsigned long fl_downgrade_time;
1013
1014        const struct file_lock_operations *fl_ops;      /* Callbacks for filesystems */
1015        const struct lock_manager_operations *fl_lmops; /* Callbacks for lockmanagers */
1016        union {
1017                struct nfs_lock_info    nfs_fl;
1018                struct nfs4_lock_info   nfs4_fl;
1019                struct {
1020                        struct list_head link;  /* link in AFS vnode's pending_locks list */
1021                        int state;              /* state of grant or error if -ve */
1022                } afs;
1023        } fl_u;
1024} __randomize_layout;
1025
1026struct file_lock_context {
1027        spinlock_t              flc_lock;
1028        struct list_head        flc_flock;
1029        struct list_head        flc_posix;
1030        struct list_head        flc_lease;
1031};
1032
1033/* The following constant reflects the upper bound of the file/locking space */
1034#ifndef OFFSET_MAX
1035#define INT_LIMIT(x)    (~((x)1 << (sizeof(x)*8 - 1)))
1036#define OFFSET_MAX      INT_LIMIT(loff_t)
1037#define OFFT_OFFSET_MAX INT_LIMIT(off_t)
1038#endif
1039
1040extern void send_sigio(struct fown_struct *fown, int fd, int band);
1041
1042/*
1043 * Return the inode to use for locking
1044 *
1045 * For overlayfs this should be the overlay inode, not the real inode returned
1046 * by file_inode().  For any other fs file_inode(filp) and locks_inode(filp) are
1047 * equal.
1048 */
1049static inline struct inode *locks_inode(const struct file *f)
1050{
1051        return f->f_path.dentry->d_inode;
1052}
1053
1054#ifdef CONFIG_FILE_LOCKING
1055extern int fcntl_getlk(struct file *, unsigned int, struct flock *);
1056extern int fcntl_setlk(unsigned int, struct file *, unsigned int,
1057                        struct flock *);
1058
1059#if BITS_PER_LONG == 32
1060extern int fcntl_getlk64(struct file *, unsigned int, struct flock64 *);
1061extern int fcntl_setlk64(unsigned int, struct file *, unsigned int,
1062                        struct flock64 *);
1063#endif
1064
1065extern int fcntl_setlease(unsigned int fd, struct file *filp, long arg);
1066extern int fcntl_getlease(struct file *filp);
1067
1068/* fs/locks.c */
1069void locks_free_lock_context(struct inode *inode);
1070void locks_free_lock(struct file_lock *fl);
1071extern void locks_init_lock(struct file_lock *);
1072extern struct file_lock * locks_alloc_lock(void);
1073extern void locks_copy_lock(struct file_lock *, struct file_lock *);
1074extern void locks_copy_conflock(struct file_lock *, struct file_lock *);
1075extern void locks_remove_posix(struct file *, fl_owner_t);
1076extern void locks_remove_file(struct file *);
1077extern void locks_release_private(struct file_lock *);
1078extern void posix_test_lock(struct file *, struct file_lock *);
1079extern int posix_lock_file(struct file *, struct file_lock *, struct file_lock *);
1080extern int posix_unblock_lock(struct file_lock *);
1081extern int vfs_test_lock(struct file *, struct file_lock *);
1082extern int vfs_lock_file(struct file *, unsigned int, struct file_lock *, struct file_lock *);
1083extern int vfs_cancel_lock(struct file *filp, struct file_lock *fl);
1084extern int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl);
1085extern int __break_lease(struct inode *inode, unsigned int flags, unsigned int type);
1086extern void lease_get_mtime(struct inode *, struct timespec *time);
1087extern int generic_setlease(struct file *, long, struct file_lock **, void **priv);
1088extern int vfs_setlease(struct file *, long, struct file_lock **, void **);
1089extern int lease_modify(struct file_lock *, int, struct list_head *);
1090struct files_struct;
1091extern void show_fd_locks(struct seq_file *f,
1092                         struct file *filp, struct files_struct *files);
1093#else /* !CONFIG_FILE_LOCKING */
1094static inline int fcntl_getlk(struct file *file, unsigned int cmd,
1095                              struct flock __user *user)
1096{
1097        return -EINVAL;
1098}
1099
1100static inline int fcntl_setlk(unsigned int fd, struct file *file,
1101                              unsigned int cmd, struct flock __user *user)
1102{
1103        return -EACCES;
1104}
1105
1106#if BITS_PER_LONG == 32
1107static inline int fcntl_getlk64(struct file *file, unsigned int cmd,
1108                                struct flock64 __user *user)
1109{
1110        return -EINVAL;
1111}
1112
1113static inline int fcntl_setlk64(unsigned int fd, struct file *file,
1114                                unsigned int cmd, struct flock64 __user *user)
1115{
1116        return -EACCES;
1117}
1118#endif
1119static inline int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1120{
1121        return -EINVAL;
1122}
1123
1124static inline int fcntl_getlease(struct file *filp)
1125{
1126        return F_UNLCK;
1127}
1128
1129static inline void
1130locks_free_lock_context(struct inode *inode)
1131{
1132}
1133
1134static inline void locks_init_lock(struct file_lock *fl)
1135{
1136        return;
1137}
1138
1139static inline void locks_copy_conflock(struct file_lock *new, struct file_lock *fl)
1140{
1141        return;
1142}
1143
1144static inline void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
1145{
1146        return;
1147}
1148
1149static inline void locks_remove_posix(struct file *filp, fl_owner_t owner)
1150{
1151        return;
1152}
1153
1154static inline void locks_remove_file(struct file *filp)
1155{
1156        return;
1157}
1158
1159static inline void posix_test_lock(struct file *filp, struct file_lock *fl)
1160{
1161        return;
1162}
1163
1164static inline int posix_lock_file(struct file *filp, struct file_lock *fl,
1165                                  struct file_lock *conflock)
1166{
1167        return -ENOLCK;
1168}
1169
1170static inline int posix_unblock_lock(struct file_lock *waiter)
1171{
1172        return -ENOENT;
1173}
1174
1175static inline int vfs_test_lock(struct file *filp, struct file_lock *fl)
1176{
1177        return 0;
1178}
1179
1180static inline int vfs_lock_file(struct file *filp, unsigned int cmd,
1181                                struct file_lock *fl, struct file_lock *conf)
1182{
1183        return -ENOLCK;
1184}
1185
1186static inline int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
1187{
1188        return 0;
1189}
1190
1191static inline int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1192{
1193        return -ENOLCK;
1194}
1195
1196static inline int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
1197{
1198        return 0;
1199}
1200
1201static inline void lease_get_mtime(struct inode *inode, struct timespec *time)
1202{
1203        return;
1204}
1205
1206static inline int generic_setlease(struct file *filp, long arg,
1207                                    struct file_lock **flp, void **priv)
1208{
1209        return -EINVAL;
1210}
1211
1212static inline int vfs_setlease(struct file *filp, long arg,
1213                               struct file_lock **lease, void **priv)
1214{
1215        return -EINVAL;
1216}
1217
1218static inline int lease_modify(struct file_lock *fl, int arg,
1219                               struct list_head *dispose)
1220{
1221        return -EINVAL;
1222}
1223
1224struct files_struct;
1225static inline void show_fd_locks(struct seq_file *f,
1226                        struct file *filp, struct files_struct *files) {}
1227#endif /* !CONFIG_FILE_LOCKING */
1228
1229static inline struct inode *file_inode(const struct file *f)
1230{
1231        return f->f_inode;
1232}
1233
1234static inline struct dentry *file_dentry(const struct file *file)
1235{
1236        return d_real(file->f_path.dentry, file_inode(file), 0);
1237}
1238
1239static inline int locks_lock_file_wait(struct file *filp, struct file_lock *fl)
1240{
1241        return locks_lock_inode_wait(locks_inode(filp), fl);
1242}
1243
1244struct fasync_struct {
1245        spinlock_t              fa_lock;
1246        int                     magic;
1247        int                     fa_fd;
1248        struct fasync_struct    *fa_next; /* singly linked list */
1249        struct file             *fa_file;
1250        struct rcu_head         fa_rcu;
1251};
1252
1253#define FASYNC_MAGIC 0x4601
1254
1255/* SMP safe fasync helpers: */
1256extern int fasync_helper(int, struct file *, int, struct fasync_struct **);
1257extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *);
1258extern int fasync_remove_entry(struct file *, struct fasync_struct **);
1259extern struct fasync_struct *fasync_alloc(void);
1260extern void fasync_free(struct fasync_struct *);
1261
1262/* can be called from interrupts */
1263extern void kill_fasync(struct fasync_struct **, int, int);
1264
1265extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force);
1266extern int f_setown(struct file *filp, unsigned long arg, int force);
1267extern void f_delown(struct file *filp);
1268extern pid_t f_getown(struct file *filp);
1269extern int send_sigurg(struct fown_struct *fown);
1270
1271struct mm_struct;
1272
1273/*
1274 *      Umount options
1275 */
1276
1277#define MNT_FORCE       0x00000001      /* Attempt to forcibily umount */
1278#define MNT_DETACH      0x00000002      /* Just detach from the tree */
1279#define MNT_EXPIRE      0x00000004      /* Mark for expiry */
1280#define UMOUNT_NOFOLLOW 0x00000008      /* Don't follow symlink on umount */
1281#define UMOUNT_UNUSED   0x80000000      /* Flag guaranteed to be unused */
1282
1283/* sb->s_iflags */
1284#define SB_I_CGROUPWB   0x00000001      /* cgroup-aware writeback enabled */
1285#define SB_I_NOEXEC     0x00000002      /* Ignore executables on this fs */
1286#define SB_I_NODEV      0x00000004      /* Ignore devices on this fs */
1287
1288/* sb->s_iflags to limit user namespace mounts */
1289#define SB_I_USERNS_VISIBLE             0x00000010 /* fstype already mounted */
1290
1291/* Possible states of 'frozen' field */
1292enum {
1293        SB_UNFROZEN = 0,                /* FS is unfrozen */
1294        SB_FREEZE_WRITE = 1,            /* Writes, dir ops, ioctls frozen */
1295        SB_FREEZE_PAGEFAULT = 2,        /* Page faults stopped as well */
1296        SB_FREEZE_FS = 3,               /* For internal FS use (e.g. to stop
1297                                         * internal threads if needed) */
1298        SB_FREEZE_COMPLETE = 4,         /* ->freeze_fs finished successfully */
1299};
1300
1301#define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1)
1302
1303struct sb_writers {
1304        int                             frozen;         /* Is sb frozen? */
1305        wait_queue_head_t               wait_unfrozen;  /* for get_super_thawed() */
1306        struct percpu_rw_semaphore      rw_sem[SB_FREEZE_LEVELS];
1307};
1308
1309struct super_block {
1310        struct list_head        s_list;         /* Keep this first */
1311        dev_t                   s_dev;          /* search index; _not_ kdev_t */
1312        unsigned char           s_blocksize_bits;
1313        unsigned long           s_blocksize;
1314        loff_t                  s_maxbytes;     /* Max file size */
1315        struct file_system_type *s_type;
1316        const struct super_operations   *s_op;
1317        const struct dquot_operations   *dq_op;
1318        const struct quotactl_ops       *s_qcop;
1319        const struct export_operations *s_export_op;
1320        unsigned long           s_flags;
1321        unsigned long           s_iflags;       /* internal SB_I_* flags */
1322        unsigned long           s_magic;
1323        struct dentry           *s_root;
1324        struct rw_semaphore     s_umount;
1325        int                     s_count;
1326        atomic_t                s_active;
1327#ifdef CONFIG_SECURITY
1328        void                    *s_security;
1329#endif
1330        const struct xattr_handler **s_xattr;
1331
1332        const struct fscrypt_operations *s_cop;
1333
1334        struct hlist_bl_head    s_anon;         /* anonymous dentries for (nfs) exporting */
1335        struct list_head        s_mounts;       /* list of mounts; _not_ for fs use */
1336        struct block_device     *s_bdev;
1337        struct backing_dev_info *s_bdi;
1338        struct mtd_info         *s_mtd;
1339        struct hlist_node       s_instances;
1340        unsigned int            s_quota_types;  /* Bitmask of supported quota types */
1341        struct quota_info       s_dquot;        /* Diskquota specific options */
1342
1343        struct sb_writers       s_writers;
1344
1345        char                    s_id[32];       /* Informational name */
1346        uuid_t                  s_uuid;         /* UUID */
1347
1348        void                    *s_fs_info;     /* Filesystem private info */
1349        unsigned int            s_max_links;
1350        fmode_t                 s_mode;
1351
1352        /* Granularity of c/m/atime in ns.
1353           Cannot be worse than a second */
1354        u32                s_time_gran;
1355
1356        /*
1357         * The next field is for VFS *only*. No filesystems have any business
1358         * even looking at it. You had been warned.
1359         */
1360        struct mutex s_vfs_rename_mutex;        /* Kludge */
1361
1362        /*
1363         * Filesystem subtype.  If non-empty the filesystem type field
1364         * in /proc/mounts will be "type.subtype"
1365         */
1366        char *s_subtype;
1367
1368        const struct dentry_operations *s_d_op; /* default d_op for dentries */
1369
1370        /*
1371         * Saved pool identifier for cleancache (-1 means none)
1372         */
1373        int cleancache_poolid;
1374
1375        struct shrinker s_shrink;       /* per-sb shrinker handle */
1376
1377        /* Number of inodes with nlink == 0 but still referenced */
1378        atomic_long_t s_remove_count;
1379
1380        /* Being remounted read-only */
1381        int s_readonly_remount;
1382
1383        /* AIO completions deferred from interrupt context */
1384        struct workqueue_struct *s_dio_done_wq;
1385        struct hlist_head s_pins;
1386
1387        /*
1388         * Owning user namespace and default context in which to
1389         * interpret filesystem uids, gids, quotas, device nodes,
1390         * xattrs and security labels.
1391         */
1392        struct user_namespace *s_user_ns;
1393
1394        /*
1395         * Keep the lru lists last in the structure so they always sit on their
1396         * own individual cachelines.
1397         */
1398        struct list_lru         s_dentry_lru ____cacheline_aligned_in_smp;
1399        struct list_lru         s_inode_lru ____cacheline_aligned_in_smp;
1400        struct rcu_head         rcu;
1401        struct work_struct      destroy_work;
1402
1403        struct mutex            s_sync_lock;    /* sync serialisation lock */
1404
1405        /*
1406         * Indicates how deep in a filesystem stack this SB is
1407         */
1408        int s_stack_depth;
1409
1410        /* s_inode_list_lock protects s_inodes */
1411        spinlock_t              s_inode_list_lock ____cacheline_aligned_in_smp;
1412        struct list_head        s_inodes;       /* all inodes */
1413
1414        spinlock_t              s_inode_wblist_lock;
1415        struct list_head        s_inodes_wb;    /* writeback inodes */
1416} __randomize_layout;
1417
1418/* Helper functions so that in most cases filesystems will
1419 * not need to deal directly with kuid_t and kgid_t and can
1420 * instead deal with the raw numeric values that are stored
1421 * in the filesystem.
1422 */
1423static inline uid_t i_uid_read(const struct inode *inode)
1424{
1425        return from_kuid(inode->i_sb->s_user_ns, inode->i_uid);
1426}
1427
1428static inline gid_t i_gid_read(const struct inode *inode)
1429{
1430        return from_kgid(inode->i_sb->s_user_ns, inode->i_gid);
1431}
1432
1433static inline void i_uid_write(struct inode *inode, uid_t uid)
1434{
1435        inode->i_uid = make_kuid(inode->i_sb->s_user_ns, uid);
1436}
1437
1438static inline void i_gid_write(struct inode *inode, gid_t gid)
1439{
1440        inode->i_gid = make_kgid(inode->i_sb->s_user_ns, gid);
1441}
1442
1443extern struct timespec current_time(struct inode *inode);
1444
1445/*
1446 * Snapshotting support.
1447 */
1448
1449void __sb_end_write(struct super_block *sb, int level);
1450int __sb_start_write(struct super_block *sb, int level, bool wait);
1451
1452#define __sb_writers_acquired(sb, lev)  \
1453        percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1454#define __sb_writers_release(sb, lev)   \
1455        percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1456
1457/**
1458 * sb_end_write - drop write access to a superblock
1459 * @sb: the super we wrote to
1460 *
1461 * Decrement number of writers to the filesystem. Wake up possible waiters
1462 * wanting to freeze the filesystem.
1463 */
1464static inline void sb_end_write(struct super_block *sb)
1465{
1466        __sb_end_write(sb, SB_FREEZE_WRITE);
1467}
1468
1469/**
1470 * sb_end_pagefault - drop write access to a superblock from a page fault
1471 * @sb: the super we wrote to
1472 *
1473 * Decrement number of processes handling write page fault to the filesystem.
1474 * Wake up possible waiters wanting to freeze the filesystem.
1475 */
1476static inline void sb_end_pagefault(struct super_block *sb)
1477{
1478        __sb_end_write(sb, SB_FREEZE_PAGEFAULT);
1479}
1480
1481/**
1482 * sb_end_intwrite - drop write access to a superblock for internal fs purposes
1483 * @sb: the super we wrote to
1484 *
1485 * Decrement fs-internal number of writers to the filesystem.  Wake up possible
1486 * waiters wanting to freeze the filesystem.
1487 */
1488static inline void sb_end_intwrite(struct super_block *sb)
1489{
1490        __sb_end_write(sb, SB_FREEZE_FS);
1491}
1492
1493/**
1494 * sb_start_write - get write access to a superblock
1495 * @sb: the super we write to
1496 *
1497 * When a process wants to write data or metadata to a file system (i.e. dirty
1498 * a page or an inode), it should embed the operation in a sb_start_write() -
1499 * sb_end_write() pair to get exclusion against file system freezing. This
1500 * function increments number of writers preventing freezing. If the file
1501 * system is already frozen, the function waits until the file system is
1502 * thawed.
1503 *
1504 * Since freeze protection behaves as a lock, users have to preserve
1505 * ordering of freeze protection and other filesystem locks. Generally,
1506 * freeze protection should be the outermost lock. In particular, we have:
1507 *
1508 * sb_start_write
1509 *   -> i_mutex                 (write path, truncate, directory ops, ...)
1510 *   -> s_umount                (freeze_super, thaw_super)
1511 */
1512static inline void sb_start_write(struct super_block *sb)
1513{
1514        __sb_start_write(sb, SB_FREEZE_WRITE, true);
1515}
1516
1517static inline int sb_start_write_trylock(struct super_block *sb)
1518{
1519        return __sb_start_write(sb, SB_FREEZE_WRITE, false);
1520}
1521
1522/**
1523 * sb_start_pagefault - get write access to a superblock from a page fault
1524 * @sb: the super we write to
1525 *
1526 * When a process starts handling write page fault, it should embed the
1527 * operation into sb_start_pagefault() - sb_end_pagefault() pair to get
1528 * exclusion against file system freezing. This is needed since the page fault
1529 * is going to dirty a page. This function increments number of running page
1530 * faults preventing freezing. If the file system is already frozen, the
1531 * function waits until the file system is thawed.
1532 *
1533 * Since page fault freeze protection behaves as a lock, users have to preserve
1534 * ordering of freeze protection and other filesystem locks. It is advised to
1535 * put sb_start_pagefault() close to mmap_sem in lock ordering. Page fault
1536 * handling code implies lock dependency:
1537 *
1538 * mmap_sem
1539 *   -> sb_start_pagefault
1540 */
1541static inline void sb_start_pagefault(struct super_block *sb)
1542{
1543        __sb_start_write(sb, SB_FREEZE_PAGEFAULT, true);
1544}
1545
1546/*
1547 * sb_start_intwrite - get write access to a superblock for internal fs purposes
1548 * @sb: the super we write to
1549 *
1550 * This is the third level of protection against filesystem freezing. It is
1551 * free for use by a filesystem. The only requirement is that it must rank
1552 * below sb_start_pagefault.
1553 *
1554 * For example filesystem can call sb_start_intwrite() when starting a
1555 * transaction which somewhat eases handling of freezing for internal sources
1556 * of filesystem changes (internal fs threads, discarding preallocation on file
1557 * close, etc.).
1558 */
1559static inline void sb_start_intwrite(struct super_block *sb)
1560{
1561        __sb_start_write(sb, SB_FREEZE_FS, true);
1562}
1563
1564
1565extern bool inode_owner_or_capable(const struct inode *inode);
1566
1567/*
1568 * VFS helper functions..
1569 */
1570extern int vfs_create(struct inode *, struct dentry *, umode_t, bool);
1571extern int vfs_mkdir(struct inode *, struct dentry *, umode_t);
1572extern int vfs_mknod(struct inode *, struct dentry *, umode_t, dev_t);
1573extern int vfs_symlink(struct inode *, struct dentry *, const char *);
1574extern int vfs_link(struct dentry *, struct inode *, struct dentry *, struct inode **);
1575extern int vfs_rmdir(struct inode *, struct dentry *);
1576extern int vfs_unlink(struct inode *, struct dentry *, struct inode **);
1577extern int vfs_rename(struct inode *, struct dentry *, struct inode *, struct dentry *, struct inode **, unsigned int);
1578extern int vfs_whiteout(struct inode *, struct dentry *);
1579
1580extern struct dentry *vfs_tmpfile(struct dentry *dentry, umode_t mode,
1581                                  int open_flag);
1582
1583/*
1584 * VFS file helper functions.
1585 */
1586extern void inode_init_owner(struct inode *inode, const struct inode *dir,
1587                        umode_t mode);
1588extern bool may_open_dev(const struct path *path);
1589/*
1590 * VFS FS_IOC_FIEMAP helper definitions.
1591 */
1592struct fiemap_extent_info {
1593        unsigned int fi_flags;          /* Flags as passed from user */
1594        unsigned int fi_extents_mapped; /* Number of mapped extents */
1595        unsigned int fi_extents_max;    /* Size of fiemap_extent array */
1596        struct fiemap_extent __user *fi_extents_start; /* Start of
1597                                                        fiemap_extent array */
1598};
1599int fiemap_fill_next_extent(struct fiemap_extent_info *info, u64 logical,
1600                            u64 phys, u64 len, u32 flags);
1601int fiemap_check_flags(struct fiemap_extent_info *fieinfo, u32 fs_flags);
1602
1603/*
1604 * File types
1605 *
1606 * NOTE! These match bits 12..15 of stat.st_mode
1607 * (ie "(i_mode >> 12) & 15").
1608 */
1609#define DT_UNKNOWN      0
1610#define DT_FIFO         1
1611#define DT_CHR          2
1612#define DT_DIR          4
1613#define DT_BLK          6
1614#define DT_REG          8
1615#define DT_LNK          10
1616#define DT_SOCK         12
1617#define DT_WHT          14
1618
1619/*
1620 * This is the "filldir" function type, used by readdir() to let
1621 * the kernel specify what kind of dirent layout it wants to have.
1622 * This allows the kernel to read directories into kernel space or
1623 * to have different dirent layouts depending on the binary type.
1624 */
1625struct dir_context;
1626typedef int (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64,
1627                         unsigned);
1628
1629struct dir_context {
1630        const filldir_t actor;
1631        loff_t pos;
1632};
1633
1634struct block_device_operations;
1635
1636/* These macros are for out of kernel modules to test that
1637 * the kernel supports the unlocked_ioctl and compat_ioctl
1638 * fields in struct file_operations. */
1639#define HAVE_COMPAT_IOCTL 1
1640#define HAVE_UNLOCKED_IOCTL 1
1641
1642/*
1643 * These flags let !MMU mmap() govern direct device mapping vs immediate
1644 * copying more easily for MAP_PRIVATE, especially for ROM filesystems.
1645 *
1646 * NOMMU_MAP_COPY:      Copy can be mapped (MAP_PRIVATE)
1647 * NOMMU_MAP_DIRECT:    Can be mapped directly (MAP_SHARED)
1648 * NOMMU_MAP_READ:      Can be mapped for reading
1649 * NOMMU_MAP_WRITE:     Can be mapped for writing
1650 * NOMMU_MAP_EXEC:      Can be mapped for execution
1651 */
1652#define NOMMU_MAP_COPY          0x00000001
1653#define NOMMU_MAP_DIRECT        0x00000008
1654#define NOMMU_MAP_READ          VM_MAYREAD
1655#define NOMMU_MAP_WRITE         VM_MAYWRITE
1656#define NOMMU_MAP_EXEC          VM_MAYEXEC
1657
1658#define NOMMU_VMFLAGS \
1659        (NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC)
1660
1661
1662struct iov_iter;
1663
1664struct file_operations {
1665        struct module *owner;
1666        loff_t (*llseek) (struct file *, loff_t, int);
1667        ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
1668        ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
1669        ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
1670        ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
1671        int (*iterate) (struct file *, struct dir_context *);
1672        int (*iterate_shared) (struct file *, struct dir_context *);
1673        unsigned int (*poll) (struct file *, struct poll_table_struct *);
1674        long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
1675        long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
1676        int (*mmap) (struct file *, struct vm_area_struct *);
1677        int (*open) (struct inode *, struct file *);
1678        int (*flush) (struct file *, fl_owner_t id);
1679        int (*release) (struct inode *, struct file *);
1680        int (*fsync) (struct file *, loff_t, loff_t, int datasync);
1681        int (*fasync) (int, struct file *, int);
1682        int (*lock) (struct file *, int, struct file_lock *);
1683        ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
1684        unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1685        int (*check_flags)(int);
1686        int (*flock) (struct file *, int, struct file_lock *);
1687        ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
1688        ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
1689        int (*setlease)(struct file *, long, struct file_lock **, void **);
1690        long (*fallocate)(struct file *file, int mode, loff_t offset,
1691                          loff_t len);
1692        void (*show_fdinfo)(struct seq_file *m, struct file *f);
1693#ifndef CONFIG_MMU
1694        unsigned (*mmap_capabilities)(struct file *);
1695#endif
1696        ssize_t (*copy_file_range)(struct file *, loff_t, struct file *,
1697                        loff_t, size_t, unsigned int);
1698        int (*clone_file_range)(struct file *, loff_t, struct file *, loff_t,
1699                        u64);
1700        ssize_t (*dedupe_file_range)(struct file *, u64, u64, struct file *,
1701                        u64);
1702} __randomize_layout;
1703
1704struct inode_operations {
1705        struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
1706        const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *);
1707        int (*permission) (struct inode *, int);
1708        struct posix_acl * (*get_acl)(struct inode *, int);
1709
1710        int (*readlink) (struct dentry *, char __user *,int);
1711
1712        int (*create) (struct inode *,struct dentry *, umode_t, bool);
1713        int (*link) (struct dentry *,struct inode *,struct dentry *);
1714        int (*unlink) (struct inode *,struct dentry *);
1715        int (*symlink) (struct inode *,struct dentry *,const char *);
1716        int (*mkdir) (struct inode *,struct dentry *,umode_t);
1717        int (*rmdir) (struct inode *,struct dentry *);
1718        int (*mknod) (struct inode *,struct dentry *,umode_t,dev_t);
1719        int (*rename) (struct inode *, struct dentry *,
1720                        struct inode *, struct dentry *, unsigned int);
1721        int (*setattr) (struct dentry *, struct iattr *);
1722        int (*getattr) (const struct path *, struct kstat *, u32, unsigned int);
1723        ssize_t (*listxattr) (struct dentry *, char *, size_t);
1724        int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start,
1725                      u64 len);
1726        int (*update_time)(struct inode *, struct timespec *, int);
1727        int (*atomic_open)(struct inode *, struct dentry *,
1728                           struct file *, unsigned open_flag,
1729                           umode_t create_mode, int *opened);
1730        int (*tmpfile) (struct inode *, struct dentry *, umode_t);
1731        int (*set_acl)(struct inode *, struct posix_acl *, int);
1732} ____cacheline_aligned;
1733
1734static inline ssize_t call_read_iter(struct file *file, struct kiocb *kio,
1735                                     struct iov_iter *iter)
1736{
1737        return file->f_op->read_iter(kio, iter);
1738}
1739
1740static inline ssize_t call_write_iter(struct file *file, struct kiocb *kio,
1741                                      struct iov_iter *iter)
1742{
1743        return file->f_op->write_iter(kio, iter);
1744}
1745
1746static inline int call_mmap(struct file *file, struct vm_area_struct *vma)
1747{
1748        return file->f_op->mmap(file, vma);
1749}
1750
1751ssize_t rw_copy_check_uvector(int type, const struct iovec __user * uvector,
1752                              unsigned long nr_segs, unsigned long fast_segs,
1753                              struct iovec *fast_pointer,
1754                              struct iovec **ret_pointer);
1755
1756extern ssize_t __vfs_read(struct file *, char __user *, size_t, loff_t *);
1757extern ssize_t __vfs_write(struct file *, const char __user *, size_t, loff_t *);
1758extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
1759extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *);
1760extern ssize_t vfs_readv(struct file *, const struct iovec __user *,
1761                unsigned long, loff_t *, int);
1762extern ssize_t vfs_writev(struct file *, const struct iovec __user *,
1763                unsigned long, loff_t *, int);
1764extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *,
1765                                   loff_t, size_t, unsigned int);
1766extern int vfs_clone_file_prep_inodes(struct inode *inode_in, loff_t pos_in,
1767                                      struct inode *inode_out, loff_t pos_out,
1768                                      u64 *len, bool is_dedupe);
1769extern int vfs_clone_file_range(struct file *file_in, loff_t pos_in,
1770                struct file *file_out, loff_t pos_out, u64 len);
1771extern int vfs_dedupe_file_range_compare(struct inode *src, loff_t srcoff,
1772                                         struct inode *dest, loff_t destoff,
1773                                         loff_t len, bool *is_same);
1774extern int vfs_dedupe_file_range(struct file *file,
1775                                 struct file_dedupe_range *same);
1776
1777struct super_operations {
1778        struct inode *(*alloc_inode)(struct super_block *sb);
1779        void (*destroy_inode)(struct inode *);
1780
1781        void (*dirty_inode) (struct inode *, int flags);
1782        int (*write_inode) (struct inode *, struct writeback_control *wbc);
1783        int (*drop_inode) (struct inode *);
1784        void (*evict_inode) (struct inode *);
1785        void (*put_super) (struct super_block *);
1786        int (*sync_fs)(struct super_block *sb, int wait);
1787        int (*freeze_super) (struct super_block *);
1788        int (*freeze_fs) (struct super_block *);
1789        int (*thaw_super) (struct super_block *);
1790        int (*unfreeze_fs) (struct super_block *);
1791        int (*statfs) (struct dentry *, struct kstatfs *);
1792        int (*remount_fs) (struct super_block *, int *, char *);
1793        void (*umount_begin) (struct super_block *);
1794
1795        int (*show_options)(struct seq_file *, struct dentry *);
1796        int (*show_devname)(struct seq_file *, struct dentry *);
1797        int (*show_path)(struct seq_file *, struct dentry *);
1798        int (*show_stats)(struct seq_file *, struct dentry *);
1799#ifdef CONFIG_QUOTA
1800        ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
1801        ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
1802        struct dquot **(*get_dquots)(struct inode *);
1803#endif
1804        int (*bdev_try_to_free_page)(struct super_block*, struct page*, gfp_t);
1805        long (*nr_cached_objects)(struct super_block *,
1806                                  struct shrink_control *);
1807        long (*free_cached_objects)(struct super_block *,
1808                                    struct shrink_control *);
1809};
1810
1811/*
1812 * Inode flags - they have no relation to superblock flags now
1813 */
1814#define S_SYNC          1       /* Writes are synced at once */
1815#define S_NOATIME       2       /* Do not update access times */
1816#define S_APPEND        4       /* Append-only file */
1817#define S_IMMUTABLE     8       /* Immutable file */
1818#define S_DEAD          16      /* removed, but still open directory */
1819#define S_NOQUOTA       32      /* Inode is not counted to quota */
1820#define S_DIRSYNC       64      /* Directory modifications are synchronous */
1821#define S_NOCMTIME      128     /* Do not update file c/mtime */
1822#define S_SWAPFILE      256     /* Do not truncate: swapon got its bmaps */
1823#define S_PRIVATE       512     /* Inode is fs-internal */
1824#define S_IMA           1024    /* Inode has an associated IMA struct */
1825#define S_AUTOMOUNT     2048    /* Automount/referral quasi-directory */
1826#define S_NOSEC         4096    /* no suid or xattr security attributes */
1827#ifdef CONFIG_FS_DAX
1828#define S_DAX           8192    /* Direct Access, avoiding the page cache */
1829#else
1830#define S_DAX           0       /* Make all the DAX code disappear */
1831#endif
1832
1833/*
1834 * Note that nosuid etc flags are inode-specific: setting some file-system
1835 * flags just means all the inodes inherit those flags by default. It might be
1836 * possible to override it selectively if you really wanted to with some
1837 * ioctl() that is not currently implemented.
1838 *
1839 * Exception: MS_RDONLY is always applied to the entire file system.
1840 *
1841 * Unfortunately, it is possible to change a filesystems flags with it mounted
1842 * with files in use.  This means that all of the inodes will not have their
1843 * i_flags updated.  Hence, i_flags no longer inherit the superblock mount
1844 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org
1845 */
1846#define __IS_FLG(inode, flg)    ((inode)->i_sb->s_flags & (flg))
1847
1848#define IS_RDONLY(inode)        ((inode)->i_sb->s_flags & MS_RDONLY)
1849#define IS_SYNC(inode)          (__IS_FLG(inode, MS_SYNCHRONOUS) || \
1850                                        ((inode)->i_flags & S_SYNC))
1851#define IS_DIRSYNC(inode)       (__IS_FLG(inode, MS_SYNCHRONOUS|MS_DIRSYNC) || \
1852                                        ((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
1853#define IS_MANDLOCK(inode)      __IS_FLG(inode, MS_MANDLOCK)
1854#define IS_NOATIME(inode)       __IS_FLG(inode, MS_RDONLY|MS_NOATIME)
1855#define IS_I_VERSION(inode)     __IS_FLG(inode, MS_I_VERSION)
1856
1857#define IS_NOQUOTA(inode)       ((inode)->i_flags & S_NOQUOTA)
1858#define IS_APPEND(inode)        ((inode)->i_flags & S_APPEND)
1859#define IS_IMMUTABLE(inode)     ((inode)->i_flags & S_IMMUTABLE)
1860#define IS_POSIXACL(inode)      __IS_FLG(inode, MS_POSIXACL)
1861
1862#define IS_DEADDIR(inode)       ((inode)->i_flags & S_DEAD)
1863#define IS_NOCMTIME(inode)      ((inode)->i_flags & S_NOCMTIME)
1864#define IS_SWAPFILE(inode)      ((inode)->i_flags & S_SWAPFILE)
1865#define IS_PRIVATE(inode)       ((inode)->i_flags & S_PRIVATE)
1866#define IS_IMA(inode)           ((inode)->i_flags & S_IMA)
1867#define IS_AUTOMOUNT(inode)     ((inode)->i_flags & S_AUTOMOUNT)
1868#define IS_NOSEC(inode)         ((inode)->i_flags & S_NOSEC)
1869#define IS_DAX(inode)           ((inode)->i_flags & S_DAX)
1870
1871#define IS_WHITEOUT(inode)      (S_ISCHR(inode->i_mode) && \
1872                                 (inode)->i_rdev == WHITEOUT_DEV)
1873
1874static inline bool HAS_UNMAPPED_ID(struct inode *inode)
1875{
1876        return !uid_valid(inode->i_uid) || !gid_valid(inode->i_gid);
1877}
1878
1879static inline enum rw_hint file_write_hint(struct file *file)
1880{
1881        if (file->f_write_hint != WRITE_LIFE_NOT_SET)
1882                return file->f_write_hint;
1883
1884        return file_inode(file)->i_write_hint;
1885}
1886
1887static inline int iocb_flags(struct file *file);
1888
1889static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp)
1890{
1891        *kiocb = (struct kiocb) {
1892                .ki_filp = filp,
1893                .ki_flags = iocb_flags(filp),
1894                .ki_hint = file_write_hint(filp),
1895        };
1896}
1897
1898/*
1899 * Inode state bits.  Protected by inode->i_lock
1900 *
1901 * Three bits determine the dirty state of the inode, I_DIRTY_SYNC,
1902 * I_DIRTY_DATASYNC and I_DIRTY_PAGES.
1903 *
1904 * Four bits define the lifetime of an inode.  Initially, inodes are I_NEW,
1905 * until that flag is cleared.  I_WILL_FREE, I_FREEING and I_CLEAR are set at
1906 * various stages of removing an inode.
1907 *
1908 * Two bits are used for locking and completion notification, I_NEW and I_SYNC.
1909 *
1910 * I_DIRTY_SYNC         Inode is dirty, but doesn't have to be written on
1911 *                      fdatasync().  i_atime is the usual cause.
1912 * I_DIRTY_DATASYNC     Data-related inode changes pending. We keep track of
1913 *                      these changes separately from I_DIRTY_SYNC so that we
1914 *                      don't have to write inode on fdatasync() when only
1915 *                      mtime has changed in it.
1916 * I_DIRTY_PAGES        Inode has dirty pages.  Inode itself may be clean.
1917 * I_NEW                Serves as both a mutex and completion notification.
1918 *                      New inodes set I_NEW.  If two processes both create
1919 *                      the same inode, one of them will release its inode and
1920 *                      wait for I_NEW to be released before returning.
1921 *                      Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can
1922 *                      also cause waiting on I_NEW, without I_NEW actually
1923 *                      being set.  find_inode() uses this to prevent returning
1924 *                      nearly-dead inodes.
1925 * I_WILL_FREE          Must be set when calling write_inode_now() if i_count
1926 *                      is zero.  I_FREEING must be set when I_WILL_FREE is
1927 *                      cleared.
1928 * I_FREEING            Set when inode is about to be freed but still has dirty
1929 *                      pages or buffers attached or the inode itself is still
1930 *                      dirty.
1931 * I_CLEAR              Added by clear_inode().  In this state the inode is
1932 *                      clean and can be destroyed.  Inode keeps I_FREEING.
1933 *
1934 *                      Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are
1935 *                      prohibited for many purposes.  iget() must wait for
1936 *                      the inode to be completely released, then create it
1937 *                      anew.  Other functions will just ignore such inodes,
1938 *                      if appropriate.  I_NEW is used for waiting.
1939 *
1940 * I_SYNC               Writeback of inode is running. The bit is set during
1941 *                      data writeback, and cleared with a wakeup on the bit
1942 *                      address once it is done. The bit is also used to pin
1943 *                      the inode in memory for flusher thread.
1944 *
1945 * I_REFERENCED         Marks the inode as recently references on the LRU list.
1946 *
1947 * I_DIO_WAKEUP         Never set.  Only used as a key for wait_on_bit().
1948 *
1949 * I_WB_SWITCH          Cgroup bdi_writeback switching in progress.  Used to
1950 *                      synchronize competing switching instances and to tell
1951 *                      wb stat updates to grab mapping->tree_lock.  See
1952 *                      inode_switch_wb_work_fn() for details.
1953 *
1954 * I_OVL_INUSE          Used by overlayfs to get exclusive ownership on upper
1955 *                      and work dirs among overlayfs mounts.
1956 *
1957 * Q: What is the difference between I_WILL_FREE and I_FREEING?
1958 */
1959#define I_DIRTY_SYNC            (1 << 0)
1960#define I_DIRTY_DATASYNC        (1 << 1)
1961#define I_DIRTY_PAGES           (1 << 2)
1962#define __I_NEW                 3
1963#define I_NEW                   (1 << __I_NEW)
1964#define I_WILL_FREE             (1 << 4)
1965#define I_FREEING               (1 << 5)
1966#define I_CLEAR                 (1 << 6)
1967#define __I_SYNC                7
1968#define I_SYNC                  (1 << __I_SYNC)
1969#define I_REFERENCED            (1 << 8)
1970#define __I_DIO_WAKEUP          9
1971#define I_DIO_WAKEUP            (1 << __I_DIO_WAKEUP)
1972#define I_LINKABLE              (1 << 10)
1973#define I_DIRTY_TIME            (1 << 11)
1974#define __I_DIRTY_TIME_EXPIRED  12
1975#define I_DIRTY_TIME_EXPIRED    (1 << __I_DIRTY_TIME_EXPIRED)
1976#define I_WB_SWITCH             (1 << 13)
1977#define I_OVL_INUSE                     (1 << 14)
1978
1979#define I_DIRTY (I_DIRTY_SYNC | I_DIRTY_DATASYNC | I_DIRTY_PAGES)
1980#define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME)
1981
1982extern void __mark_inode_dirty(struct inode *, int);
1983static inline void mark_inode_dirty(struct inode *inode)
1984{
1985        __mark_inode_dirty(inode, I_DIRTY);
1986}
1987
1988static inline void mark_inode_dirty_sync(struct inode *inode)
1989{
1990        __mark_inode_dirty(inode, I_DIRTY_SYNC);
1991}
1992
1993extern void inc_nlink(struct inode *inode);
1994extern void drop_nlink(struct inode *inode);
1995extern void clear_nlink(struct inode *inode);
1996extern void set_nlink(struct inode *inode, unsigned int nlink);
1997
1998static inline void inode_inc_link_count(struct inode *inode)
1999{
2000        inc_nlink(inode);
2001        mark_inode_dirty(inode);
2002}
2003
2004static inline void inode_dec_link_count(struct inode *inode)
2005{
2006        drop_nlink(inode);
2007        mark_inode_dirty(inode);
2008}
2009
2010/**
2011 * inode_inc_iversion - increments i_version
2012 * @inode: inode that need to be updated
2013 *
2014 * Every time the inode is modified, the i_version field will be incremented.
2015 * The filesystem has to be mounted with i_version flag
2016 */
2017
2018static inline void inode_inc_iversion(struct inode *inode)
2019{
2020       spin_lock(&inode->i_lock);
2021       inode->i_version++;
2022       spin_unlock(&inode->i_lock);
2023}
2024
2025enum file_time_flags {
2026        S_ATIME = 1,
2027        S_MTIME = 2,
2028        S_CTIME = 4,
2029        S_VERSION = 8,
2030};
2031
2032extern void touch_atime(const struct path *);
2033static inline void file_accessed(struct file *file)
2034{
2035        if (!(file->f_flags & O_NOATIME))
2036                touch_atime(&file->f_path);
2037}
2038
2039int sync_inode(struct inode *inode, struct writeback_control *wbc);
2040int sync_inode_metadata(struct inode *inode, int wait);
2041
2042struct file_system_type {
2043        const char *name;
2044        int fs_flags;
2045#define FS_REQUIRES_DEV         1 
2046#define FS_BINARY_MOUNTDATA     2
2047#define FS_HAS_SUBTYPE          4
2048#define FS_USERNS_MOUNT         8       /* Can be mounted by userns root */
2049#define FS_RENAME_DOES_D_MOVE   32768   /* FS will handle d_move() during rename() internally. */
2050        struct dentry *(*mount) (struct file_system_type *, int,
2051                       const char *, void *);
2052        void (*kill_sb) (struct super_block *);
2053        struct module *owner;
2054        struct file_system_type * next;
2055        struct hlist_head fs_supers;
2056
2057        struct lock_class_key s_lock_key;
2058        struct lock_class_key s_umount_key;
2059        struct lock_class_key s_vfs_rename_key;
2060        struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
2061
2062        struct lock_class_key i_lock_key;
2063        struct lock_class_key i_mutex_key;
2064        struct lock_class_key i_mutex_dir_key;
2065};
2066
2067#define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME)
2068
2069extern struct dentry *mount_ns(struct file_system_type *fs_type,
2070        int flags, void *data, void *ns, struct user_namespace *user_ns,
2071        int (*fill_super)(struct super_block *, void *, int));
2072extern struct dentry *mount_bdev(struct file_system_type *fs_type,
2073        int flags, const char *dev_name, void *data,
2074        int (*fill_super)(struct super_block *, void *, int));
2075extern struct dentry *mount_single(struct file_system_type *fs_type,
2076        int flags, void *data,
2077        int (*fill_super)(struct super_block *, void *, int));
2078extern struct dentry *mount_nodev(struct file_system_type *fs_type,
2079        int flags, void *data,
2080        int (*fill_super)(struct super_block *, void *, int));
2081extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path);
2082void generic_shutdown_super(struct super_block *sb);
2083void kill_block_super(struct super_block *sb);
2084void kill_anon_super(struct super_block *sb);
2085void kill_litter_super(struct super_block *sb);
2086void deactivate_super(struct super_block *sb);
2087void deactivate_locked_super(struct super_block *sb);
2088int set_anon_super(struct super_block *s, void *data);
2089int get_anon_bdev(dev_t *);
2090void free_anon_bdev(dev_t);
2091struct super_block *sget_userns(struct file_system_type *type,
2092                        int (*test)(struct super_block *,void *),
2093                        int (*set)(struct super_block *,void *),
2094                        int flags, struct user_namespace *user_ns,
2095                        void *data);
2096struct super_block *sget(struct file_system_type *type,
2097                        int (*test)(struct super_block *,void *),
2098                        int (*set)(struct super_block *,void *),
2099                        int flags, void *data);
2100extern struct dentry *mount_pseudo_xattr(struct file_system_type *, char *,
2101                                         const struct super_operations *ops,
2102                                         const struct xattr_handler **xattr,
2103                                         const struct dentry_operations *dops,
2104                                         unsigned long);
2105
2106static inline struct dentry *
2107mount_pseudo(struct file_system_type *fs_type, char *name,
2108             const struct super_operations *ops,
2109             const struct dentry_operations *dops, unsigned long magic)
2110{
2111        return mount_pseudo_xattr(fs_type, name, ops, NULL, dops, magic);
2112}
2113
2114/* Alas, no aliases. Too much hassle with bringing module.h everywhere */
2115#define fops_get(fops) \
2116        (((fops) && try_module_get((fops)->owner) ? (fops) : NULL))
2117#define fops_put(fops) \
2118        do { if (fops) module_put((fops)->owner); } while(0)
2119/*
2120 * This one is to be used *ONLY* from ->open() instances.
2121 * fops must be non-NULL, pinned down *and* module dependencies
2122 * should be sufficient to pin the caller down as well.
2123 */
2124#define replace_fops(f, fops) \
2125        do {    \
2126                struct file *__file = (f); \
2127                fops_put(__file->f_op); \
2128                BUG_ON(!(__file->f_op = (fops))); \
2129        } while(0)
2130
2131extern int register_filesystem(struct file_system_type *);
2132extern int unregister_filesystem(struct file_system_type *);
2133extern struct vfsmount *kern_mount_data(struct file_system_type *, void *data);
2134#define kern_mount(type) kern_mount_data(type, NULL)
2135extern void kern_unmount(struct vfsmount *mnt);
2136extern int may_umount_tree(struct vfsmount *);
2137extern int may_umount(struct vfsmount *);
2138extern long do_mount(const char *, const char __user *,
2139                     const char *, unsigned long, void *);
2140extern struct vfsmount *collect_mounts(const struct path *);
2141extern void drop_collected_mounts(struct vfsmount *);
2142extern int iterate_mounts(int (*)(struct vfsmount *, void *), void *,
2143                          struct vfsmount *);
2144extern int vfs_statfs(const struct path *, struct kstatfs *);
2145extern int user_statfs(const char __user *, struct kstatfs *);
2146extern int fd_statfs(int, struct kstatfs *);
2147extern int vfs_ustat(dev_t, struct kstatfs *);
2148extern int freeze_super(struct super_block *super);
2149extern int thaw_super(struct super_block *super);
2150extern bool our_mnt(struct vfsmount *mnt);
2151extern __printf(2, 3)
2152int super_setup_bdi_name(struct super_block *sb, char *fmt, ...);
2153extern int super_setup_bdi(struct super_block *sb);
2154
2155extern int current_umask(void);
2156
2157extern void ihold(struct inode * inode);
2158extern void iput(struct inode *);
2159extern int generic_update_time(struct inode *, struct timespec *, int);
2160
2161/* /sys/fs */
2162extern struct kobject *fs_kobj;
2163
2164#define MAX_RW_COUNT (INT_MAX & PAGE_MASK)
2165
2166#ifdef CONFIG_MANDATORY_FILE_LOCKING
2167extern int locks_mandatory_locked(struct file *);
2168extern int locks_mandatory_area(struct inode *, struct file *, loff_t, loff_t, unsigned char);
2169
2170/*
2171 * Candidates for mandatory locking have the setgid bit set
2172 * but no group execute bit -  an otherwise meaningless combination.
2173 */
2174
2175static inline int __mandatory_lock(struct inode *ino)
2176{
2177        return (ino->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID;
2178}
2179
2180/*
2181 * ... and these candidates should be on MS_MANDLOCK mounted fs,
2182 * otherwise these will be advisory locks
2183 */
2184
2185static inline int mandatory_lock(struct inode *ino)
2186{
2187        return IS_MANDLOCK(ino) && __mandatory_lock(ino);
2188}
2189
2190static inline int locks_verify_locked(struct file *file)
2191{
2192        if (mandatory_lock(locks_inode(file)))
2193                return locks_mandatory_locked(file);
2194        return 0;
2195}
2196
2197static inline int locks_verify_truncate(struct inode *inode,
2198                                    struct file *f,
2199                                    loff_t size)
2200{
2201        if (!inode->i_flctx || !mandatory_lock(inode))
2202                return 0;
2203
2204        if (size < inode->i_size) {
2205                return locks_mandatory_area(inode, f, size, inode->i_size - 1,
2206                                F_WRLCK);
2207        } else {
2208                return locks_mandatory_area(inode, f, inode->i_size, size - 1,
2209                                F_WRLCK);
2210        }
2211}
2212
2213#else /* !CONFIG_MANDATORY_FILE_LOCKING */
2214
2215static inline int locks_mandatory_locked(struct file *file)
2216{
2217        return 0;
2218}
2219
2220static inline int locks_mandatory_area(struct inode *inode, struct file *filp,
2221                                       loff_t start, loff_t end, unsigned char type)
2222{
2223        return 0;
2224}
2225
2226static inline int __mandatory_lock(struct inode *inode)
2227{
2228        return 0;
2229}
2230
2231static inline int mandatory_lock(struct inode *inode)
2232{
2233        return 0;
2234}
2235
2236static inline int locks_verify_locked(struct file *file)
2237{
2238        return 0;
2239}
2240
2241static inline int locks_verify_truncate(struct inode *inode, struct file *filp,
2242                                        size_t size)
2243{
2244        return 0;
2245}
2246
2247#endif /* CONFIG_MANDATORY_FILE_LOCKING */
2248
2249
2250#ifdef CONFIG_FILE_LOCKING
2251static inline int break_lease(struct inode *inode, unsigned int mode)
2252{
2253        /*
2254         * Since this check is lockless, we must ensure that any refcounts
2255         * taken are done before checking i_flctx->flc_lease. Otherwise, we
2256         * could end up racing with tasks trying to set a new lease on this
2257         * file.
2258         */
2259        smp_mb();
2260        if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease))
2261                return __break_lease(inode, mode, FL_LEASE);
2262        return 0;
2263}
2264
2265static inline int break_deleg(struct inode *inode, unsigned int mode)
2266{
2267        /*
2268         * Since this check is lockless, we must ensure that any refcounts
2269         * taken are done before checking i_flctx->flc_lease. Otherwise, we
2270         * could end up racing with tasks trying to set a new lease on this
2271         * file.
2272         */
2273        smp_mb();
2274        if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease))
2275                return __break_lease(inode, mode, FL_DELEG);
2276        return 0;
2277}
2278
2279static inline int try_break_deleg(struct inode *inode, struct inode **delegated_inode)
2280{
2281        int ret;
2282
2283        ret = break_deleg(inode, O_WRONLY|O_NONBLOCK);
2284        if (ret == -EWOULDBLOCK && delegated_inode) {
2285                *delegated_inode = inode;
2286                ihold(inode);
2287        }
2288        return ret;
2289}
2290
2291static inline int break_deleg_wait(struct inode **delegated_inode)
2292{
2293        int ret;
2294
2295        ret = break_deleg(*delegated_inode, O_WRONLY);
2296        iput(*delegated_inode);
2297        *delegated_inode = NULL;
2298        return ret;
2299}
2300
2301static inline int break_layout(struct inode *inode, bool wait)
2302{
2303        smp_mb();
2304        if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease))
2305                return __break_lease(inode,
2306                                wait ? O_WRONLY : O_WRONLY | O_NONBLOCK,
2307                                FL_LAYOUT);
2308        return 0;
2309}
2310
2311#else /* !CONFIG_FILE_LOCKING */
2312static inline int break_lease(struct inode *inode, unsigned int mode)
2313{
2314        return 0;
2315}
2316
2317static inline int break_deleg(struct inode *inode, unsigned int mode)
2318{
2319        return 0;
2320}
2321
2322static inline int try_break_deleg(struct inode *inode, struct inode **delegated_inode)
2323{
2324        return 0;
2325}
2326
2327static inline int break_deleg_wait(struct inode **delegated_inode)
2328{
2329        BUG();
2330        return 0;
2331}
2332
2333static inline int break_layout(struct inode *inode, bool wait)
2334{
2335        return 0;
2336}
2337
2338#endif /* CONFIG_FILE_LOCKING */
2339
2340/* fs/open.c */
2341struct audit_names;
2342struct filename {
2343        const char              *name;  /* pointer to actual string */
2344        const __user char       *uptr;  /* original userland pointer */
2345        struct audit_names      *aname;
2346        int                     refcnt;
2347        const char              iname[];
2348};
2349
2350extern long vfs_truncate(const struct path *, loff_t);
2351extern int do_truncate(struct dentry *, loff_t start, unsigned int time_attrs,
2352                       struct file *filp);
2353extern int vfs_fallocate(struct file *file, int mode, loff_t offset,
2354                        loff_t len);
2355extern long do_sys_open(int dfd, const char __user *filename, int flags,
2356                        umode_t mode);
2357extern struct file *file_open_name(struct filename *, int, umode_t);
2358extern struct file *filp_open(const char *, int, umode_t);
2359extern struct file *file_open_root(struct dentry *, struct vfsmount *,
2360                                   const char *, int, umode_t);
2361extern struct file * dentry_open(const struct path *, int, const struct cred *);
2362extern int filp_close(struct file *, fl_owner_t id);
2363
2364extern struct filename *getname_flags(const char __user *, int, int *);
2365extern struct filename *getname(const char __user *);
2366extern struct filename *getname_kernel(const char *);
2367extern void putname(struct filename *name);
2368
2369enum {
2370        FILE_CREATED = 1,
2371        FILE_OPENED = 2
2372};
2373extern int finish_open(struct file *file, struct dentry *dentry,
2374                        int (*open)(struct inode *, struct file *),
2375                        int *opened);
2376extern int finish_no_open(struct file *file, struct dentry *dentry);
2377
2378/* fs/ioctl.c */
2379
2380extern int ioctl_preallocate(struct file *filp, void __user *argp);
2381
2382/* fs/dcache.c */
2383extern void __init vfs_caches_init_early(void);
2384extern void __init vfs_caches_init(void);
2385
2386extern struct kmem_cache *names_cachep;
2387
2388#define __getname()             kmem_cache_alloc(names_cachep, GFP_KERNEL)
2389#define __putname(name)         kmem_cache_free(names_cachep, (void *)(name))
2390
2391#ifdef CONFIG_BLOCK
2392extern int register_blkdev(unsigned int, const char *);
2393extern void unregister_blkdev(unsigned int, const char *);
2394extern void bdev_unhash_inode(dev_t dev);
2395extern struct block_device *bdget(dev_t);
2396extern struct block_device *bdgrab(struct block_device *bdev);
2397extern void bd_set_size(struct block_device *, loff_t size);
2398extern void bd_forget(struct inode *inode);
2399extern void bdput(struct block_device *);
2400extern void invalidate_bdev(struct block_device *);
2401extern void iterate_bdevs(void (*)(struct block_device *, void *), void *);
2402extern int sync_blockdev(struct block_device *bdev);
2403extern void kill_bdev(struct block_device *);
2404extern struct super_block *freeze_bdev(struct block_device *);
2405extern void emergency_thaw_all(void);
2406extern int thaw_bdev(struct block_device *bdev, struct super_block *sb);
2407extern int fsync_bdev(struct block_device *);
2408
2409extern struct super_block *blockdev_superblock;
2410
2411static inline bool sb_is_blkdev_sb(struct super_block *sb)
2412{
2413        return sb == blockdev_superblock;
2414}
2415#else
2416static inline void bd_forget(struct inode *inode) {}
2417static inline int sync_blockdev(struct block_device *bdev) { return 0; }
2418static inline void kill_bdev(struct block_device *bdev) {}
2419static inline void invalidate_bdev(struct block_device *bdev) {}
2420
2421static inline struct super_block *freeze_bdev(struct block_device *sb)
2422{
2423        return NULL;
2424}
2425
2426static inline int thaw_bdev(struct block_device *bdev, struct super_block *sb)
2427{
2428        return 0;
2429}
2430
2431static inline void iterate_bdevs(void (*f)(struct block_device *, void *), void *arg)
2432{
2433}
2434
2435static inline bool sb_is_blkdev_sb(struct super_block *sb)
2436{
2437        return false;
2438}
2439#endif
2440extern int sync_filesystem(struct super_block *);
2441extern const struct file_operations def_blk_fops;
2442extern const struct file_operations def_chr_fops;
2443#ifdef CONFIG_BLOCK
2444extern int ioctl_by_bdev(struct block_device *, unsigned, unsigned long);
2445extern int blkdev_ioctl(struct block_device *, fmode_t, unsigned, unsigned long);
2446extern long compat_blkdev_ioctl(struct file *, unsigned, unsigned long);
2447extern int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder);
2448extern struct block_device *blkdev_get_by_path(const char *path, fmode_t mode,
2449                                               void *holder);
2450extern struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode,
2451                                              void *holder);
2452extern void blkdev_put(struct block_device *bdev, fmode_t mode);
2453extern int __blkdev_reread_part(struct block_device *bdev);
2454extern int blkdev_reread_part(struct block_device *bdev);
2455
2456#ifdef CONFIG_SYSFS
2457extern int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk);
2458extern void bd_unlink_disk_holder(struct block_device *bdev,
2459                                  struct gendisk *disk);
2460#else
2461static inline int bd_link_disk_holder(struct block_device *bdev,
2462                                      struct gendisk *disk)
2463{
2464        return 0;
2465}
2466static inline void bd_unlink_disk_holder(struct block_device *bdev,
2467                                         struct gendisk *disk)
2468{
2469}
2470#endif
2471#endif
2472
2473/* fs/char_dev.c */
2474#define CHRDEV_MAJOR_HASH_SIZE  255
2475/* Marks the bottom of the first segment of free char majors */
2476#define CHRDEV_MAJOR_DYN_END 234
2477extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *);
2478extern int register_chrdev_region(dev_t, unsigned, const char *);
2479extern int __register_chrdev(unsigned int major, unsigned int baseminor,
2480                             unsigned int count, const char *name,
2481                             const struct file_operations *fops);
2482extern void __unregister_chrdev(unsigned int major, unsigned int baseminor,
2483                                unsigned int count, const char *name);
2484extern void unregister_chrdev_region(dev_t, unsigned);
2485extern void chrdev_show(struct seq_file *,off_t);
2486
2487static inline int register_chrdev(unsigned int major, const char *name,
2488                                  const struct file_operations *fops)
2489{
2490        return __register_chrdev(major, 0, 256, name, fops);
2491}
2492
2493static inline void unregister_chrdev(unsigned int major, const char *name)
2494{
2495        __unregister_chrdev(major, 0, 256, name);
2496}
2497
2498/* fs/block_dev.c */
2499#define BDEVNAME_SIZE   32      /* Largest string for a blockdev identifier */
2500#define BDEVT_SIZE      10      /* Largest string for MAJ:MIN for blkdev */
2501
2502#ifdef CONFIG_BLOCK
2503#define BLKDEV_MAJOR_HASH_SIZE  255
2504extern const char *__bdevname(dev_t, char *buffer);
2505extern const char *bdevname(struct block_device *bdev, char *buffer);
2506extern struct block_device *lookup_bdev(const char *);
2507extern void blkdev_show(struct seq_file *,off_t);
2508
2509#else
2510#define BLKDEV_MAJOR_HASH_SIZE  0
2511#endif
2512
2513extern void init_special_inode(struct inode *, umode_t, dev_t);
2514
2515/* Invalid inode operations -- fs/bad_inode.c */
2516extern void make_bad_inode(struct inode *);
2517extern bool is_bad_inode(struct inode *);
2518
2519#ifdef CONFIG_BLOCK
2520extern void check_disk_size_change(struct gendisk *disk,
2521                                   struct block_device *bdev);
2522extern int revalidate_disk(struct gendisk *);
2523extern int check_disk_change(struct block_device *);
2524extern int __invalidate_device(struct block_device *, bool);
2525extern int invalidate_partition(struct gendisk *, int);
2526#endif
2527unsigned long invalidate_mapping_pages(struct address_space *mapping,
2528                                        pgoff_t start, pgoff_t end);
2529
2530static inline void invalidate_remote_inode(struct inode *inode)
2531{
2532        if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
2533            S_ISLNK(inode->i_mode))
2534                invalidate_mapping_pages(inode->i_mapping, 0, -1);
2535}
2536extern int invalidate_inode_pages2(struct address_space *mapping);
2537extern int invalidate_inode_pages2_range(struct address_space *mapping,
2538                                         pgoff_t start, pgoff_t end);
2539extern int write_inode_now(struct inode *, int);
2540extern int filemap_fdatawrite(struct address_space *);
2541extern int filemap_flush(struct address_space *);
2542extern int filemap_fdatawait(struct address_space *);
2543extern int filemap_fdatawait_keep_errors(struct address_space *mapping);
2544extern int filemap_fdatawait_range(struct address_space *, loff_t lstart,
2545                                   loff_t lend);
2546extern bool filemap_range_has_page(struct address_space *, loff_t lstart,
2547                                  loff_t lend);
2548extern int filemap_write_and_wait(struct address_space *mapping);
2549extern int filemap_write_and_wait_range(struct address_space *mapping,
2550                                        loff_t lstart, loff_t lend);
2551extern int __filemap_fdatawrite_range(struct address_space *mapping,
2552                                loff_t start, loff_t end, int sync_mode);
2553extern int filemap_fdatawrite_range(struct address_space *mapping,
2554                                loff_t start, loff_t end);
2555extern int filemap_check_errors(struct address_space *mapping);
2556
2557extern void __filemap_set_wb_err(struct address_space *mapping, int err);
2558extern int __must_check file_check_and_advance_wb_err(struct file *file);
2559extern int __must_check file_write_and_wait_range(struct file *file,
2560                                                loff_t start, loff_t end);
2561
2562/**
2563 * filemap_set_wb_err - set a writeback error on an address_space
2564 * @mapping: mapping in which to set writeback error
2565 * @err: error to be set in mapping
2566 *
2567 * When writeback fails in some way, we must record that error so that
2568 * userspace can be informed when fsync and the like are called.  We endeavor
2569 * to report errors on any file that was open at the time of the error.  Some
2570 * internal callers also need to know when writeback errors have occurred.
2571 *
2572 * When a writeback error occurs, most filesystems will want to call
2573 * filemap_set_wb_err to record the error in the mapping so that it will be
2574 * automatically reported whenever fsync is called on the file.
2575 *
2576 * FIXME: mention FS_* flag here?
2577 */
2578static inline void filemap_set_wb_err(struct address_space *mapping, int err)
2579{
2580        /* Fastpath for common case of no error */
2581        if (unlikely(err))
2582                __filemap_set_wb_err(mapping, err);
2583}
2584
2585/**
2586 * filemap_check_wb_error - has an error occurred since the mark was sampled?
2587 * @mapping: mapping to check for writeback errors
2588 * @since: previously-sampled errseq_t
2589 *
2590 * Grab the errseq_t value from the mapping, and see if it has changed "since"
2591 * the given value was sampled.
2592 *
2593 * If it has then report the latest error set, otherwise return 0.
2594 */
2595static inline int filemap_check_wb_err(struct address_space *mapping,
2596                                        errseq_t since)
2597{
2598        return errseq_check(&mapping->wb_err, since);
2599}
2600
2601/**
2602 * filemap_sample_wb_err - sample the current errseq_t to test for later errors
2603 * @mapping: mapping to be sampled
2604 *
2605 * Writeback errors are always reported relative to a particular sample point
2606 * in the past. This function provides those sample points.
2607 */
2608static inline errseq_t filemap_sample_wb_err(struct address_space *mapping)
2609{
2610        return errseq_sample(&mapping->wb_err);
2611}
2612
2613extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end,
2614                           int datasync);
2615extern int vfs_fsync(struct file *file, int datasync);
2616
2617/*
2618 * Sync the bytes written if this was a synchronous write.  Expect ki_pos
2619 * to already be updated for the write, and will return either the amount
2620 * of bytes passed in, or an error if syncing the file failed.
2621 */
2622static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count)
2623{
2624        if (iocb->ki_flags & IOCB_DSYNC) {
2625                int ret = vfs_fsync_range(iocb->ki_filp,
2626                                iocb->ki_pos - count, iocb->ki_pos - 1,
2627                                (iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
2628                if (ret)
2629                        return ret;
2630        }
2631
2632        return count;
2633}
2634
2635extern void emergency_sync(void);
2636extern void emergency_remount(void);
2637#ifdef CONFIG_BLOCK
2638extern sector_t bmap(struct inode *, sector_t);
2639#endif
2640extern int notify_change(struct dentry *, struct iattr *, struct inode **);
2641extern int inode_permission(struct inode *, int);
2642extern int __inode_permission(struct inode *, int);
2643extern int generic_permission(struct inode *, int);
2644extern int __check_sticky(struct inode *dir, struct inode *inode);
2645
2646static inline bool execute_ok(struct inode *inode)
2647{
2648        return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode);
2649}
2650
2651static inline void file_start_write(struct file *file)
2652{
2653        if (!S_ISREG(file_inode(file)->i_mode))
2654                return;
2655        __sb_start_write(file_inode(file)->i_sb, SB_FREEZE_WRITE, true);
2656}
2657
2658static inline bool file_start_write_trylock(struct file *file)
2659{
2660        if (!S_ISREG(file_inode(file)->i_mode))
2661                return true;
2662        return __sb_start_write(file_inode(file)->i_sb, SB_FREEZE_WRITE, false);
2663}
2664
2665static inline void file_end_write(struct file *file)
2666{
2667        if (!S_ISREG(file_inode(file)->i_mode))
2668                return;
2669        __sb_end_write(file_inode(file)->i_sb, SB_FREEZE_WRITE);
2670}
2671
2672static inline int do_clone_file_range(struct file *file_in, loff_t pos_in,
2673                                      struct file *file_out, loff_t pos_out,
2674                                      u64 len)
2675{
2676        int ret;
2677
2678        file_start_write(file_out);
2679        ret = vfs_clone_file_range(file_in, pos_in, file_out, pos_out, len);
2680        file_end_write(file_out);
2681
2682        return ret;
2683}
2684
2685/*
2686 * get_write_access() gets write permission for a file.
2687 * put_write_access() releases this write permission.
2688 * This is used for regular files.
2689 * We cannot support write (and maybe mmap read-write shared) accesses and
2690 * MAP_DENYWRITE mmappings simultaneously. The i_writecount field of an inode
2691 * can have the following values:
2692 * 0: no writers, no VM_DENYWRITE mappings
2693 * < 0: (-i_writecount) vm_area_structs with VM_DENYWRITE set exist
2694 * > 0: (i_writecount) users are writing to the file.
2695 *
2696 * Normally we operate on that counter with atomic_{inc,dec} and it's safe
2697 * except for the cases where we don't hold i_writecount yet. Then we need to
2698 * use {get,deny}_write_access() - these functions check the sign and refuse
2699 * to do the change if sign is wrong.
2700 */
2701static inline int get_write_access(struct inode *inode)
2702{
2703        return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY;
2704}
2705static inline int deny_write_access(struct file *file)
2706{
2707        struct inode *inode = file_inode(file);
2708        return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY;
2709}
2710static inline void put_write_access(struct inode * inode)
2711{
2712        atomic_dec(&inode->i_writecount);
2713}
2714static inline void allow_write_access(struct file *file)
2715{
2716        if (file)
2717                atomic_inc(&file_inode(file)->i_writecount);
2718}
2719static inline bool inode_is_open_for_write(const struct inode *inode)
2720{
2721        return atomic_read(&inode->i_writecount) > 0;
2722}
2723
2724#ifdef CONFIG_IMA
2725static inline void i_readcount_dec(struct inode *inode)
2726{
2727        BUG_ON(!atomic_read(&inode->i_readcount));
2728        atomic_dec(&inode->i_readcount);
2729}
2730static inline void i_readcount_inc(struct inode *inode)
2731{
2732        atomic_inc(&inode->i_readcount);
2733}
2734#else
2735static inline void i_readcount_dec(struct inode *inode)
2736{
2737        return;
2738}
2739static inline void i_readcount_inc(struct inode *inode)
2740{
2741        return;
2742}
2743#endif
2744extern int do_pipe_flags(int *, int);
2745
2746#define __kernel_read_file_id(id) \
2747        id(UNKNOWN, unknown)            \
2748        id(FIRMWARE, firmware)          \
2749        id(FIRMWARE_PREALLOC_BUFFER, firmware)  \
2750        id(MODULE, kernel-module)               \
2751        id(KEXEC_IMAGE, kexec-image)            \
2752        id(KEXEC_INITRAMFS, kexec-initramfs)    \
2753        id(POLICY, security-policy)             \
2754        id(MAX_ID, )
2755
2756#define __fid_enumify(ENUM, dummy) READING_ ## ENUM,
2757#define __fid_stringify(dummy, str) #str,
2758
2759enum kernel_read_file_id {
2760        __kernel_read_file_id(__fid_enumify)
2761};
2762
2763static const char * const kernel_read_file_str[] = {
2764        __kernel_read_file_id(__fid_stringify)
2765};
2766
2767static inline const char *kernel_read_file_id_str(enum kernel_read_file_id id)
2768{
2769        if ((unsigned)id >= READING_MAX_ID)
2770                return kernel_read_file_str[READING_UNKNOWN];
2771
2772        return kernel_read_file_str[id];
2773}
2774
2775extern int kernel_read(struct file *, loff_t, char *, unsigned long);
2776extern int kernel_read_file(struct file *, void **, loff_t *, loff_t,
2777                            enum kernel_read_file_id);
2778extern int kernel_read_file_from_path(char *, void **, loff_t *, loff_t,
2779                                      enum kernel_read_file_id);
2780extern int kernel_read_file_from_fd(int, void **, loff_t *, loff_t,
2781                                    enum kernel_read_file_id);
2782extern ssize_t kernel_write(struct file *, const char *, size_t, loff_t);
2783extern ssize_t __kernel_write(struct file *, const char *, size_t, loff_t *);
2784extern struct file * open_exec(const char *);
2785 
2786/* fs/dcache.c -- generic fs support functions */
2787extern bool is_subdir(struct dentry *, struct dentry *);
2788extern bool path_is_under(const struct path *, const struct path *);
2789
2790extern char *file_path(struct file *, char *, int);
2791
2792#include <linux/err.h>
2793
2794/* needed for stackable file system support */
2795extern loff_t default_llseek(struct file *file, loff_t offset, int whence);
2796
2797extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence);
2798
2799extern int inode_init_always(struct super_block *, struct inode *);
2800extern void inode_init_once(struct inode *);
2801extern void address_space_init_once(struct address_space *mapping);
2802extern struct inode * igrab(struct inode *);
2803extern ino_t iunique(struct super_block *, ino_t);
2804extern int inode_needs_sync(struct inode *inode);
2805extern int generic_delete_inode(struct inode *inode);
2806static inline int generic_drop_inode(struct inode *inode)
2807{
2808        return !inode->i_nlink || inode_unhashed(inode);
2809}
2810
2811extern struct inode *ilookup5_nowait(struct super_block *sb,
2812                unsigned long hashval, int (*test)(struct inode *, void *),
2813                void *data);
2814extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
2815                int (*test)(struct inode *, void *), void *data);
2816extern struct inode *ilookup(struct super_block *sb, unsigned long ino);
2817
2818extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *);
2819extern struct inode * iget_locked(struct super_block *, unsigned long);
2820extern struct inode *find_inode_nowait(struct super_block *,
2821                                       unsigned long,
2822                                       int (*match)(struct inode *,
2823                                                    unsigned long, void *),
2824                                       void *data);
2825extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *);
2826extern int insert_inode_locked(struct inode *);
2827#ifdef CONFIG_DEBUG_LOCK_ALLOC
2828extern void lockdep_annotate_inode_mutex_key(struct inode *inode);
2829#else
2830static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { };
2831#endif
2832extern void unlock_new_inode(struct inode *);
2833extern unsigned int get_next_ino(void);
2834
2835extern void __iget(struct inode * inode);
2836extern void iget_failed(struct inode *);
2837extern void clear_inode(struct inode *);
2838extern void __destroy_inode(struct inode *);
2839extern struct inode *new_inode_pseudo(struct super_block *sb);
2840extern struct inode *new_inode(struct super_block *sb);
2841extern void free_inode_nonrcu(struct inode *inode);
2842extern int should_remove_suid(struct dentry *);
2843extern int file_remove_privs(struct file *);
2844
2845extern void __insert_inode_hash(struct inode *, unsigned long hashval);
2846static inline void insert_inode_hash(struct inode *inode)
2847{
2848        __insert_inode_hash(inode, inode->i_ino);
2849}
2850
2851extern void __remove_inode_hash(struct inode *);
2852static inline void remove_inode_hash(struct inode *inode)
2853{
2854        if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash))
2855                __remove_inode_hash(inode);
2856}
2857
2858extern void inode_sb_list_add(struct inode *inode);
2859
2860#ifdef CONFIG_BLOCK
2861extern int bdev_read_only(struct block_device *);
2862#endif
2863extern int set_blocksize(struct block_device *, int);
2864extern int sb_set_blocksize(struct super_block *, int);
2865extern int sb_min_blocksize(struct super_block *, int);
2866
2867extern int generic_file_mmap(struct file *, struct vm_area_struct *);
2868extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *);
2869extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *);
2870extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *);
2871extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *);
2872extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *);
2873extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *);
2874extern ssize_t generic_perform_write(struct file *, struct iov_iter *, loff_t);
2875
2876ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos,
2877                int flags);
2878ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos,
2879                int flags);
2880
2881/* fs/block_dev.c */
2882extern ssize_t blkdev_read_iter(struct kiocb *iocb, struct iov_iter *to);
2883extern ssize_t blkdev_write_iter(struct kiocb *iocb, struct iov_iter *from);
2884extern int blkdev_fsync(struct file *filp, loff_t start, loff_t end,
2885                        int datasync);
2886extern void block_sync_page(struct page *page);
2887
2888/* fs/splice.c */
2889extern ssize_t generic_file_splice_read(struct file *, loff_t *,
2890                struct pipe_inode_info *, size_t, unsigned int);
2891extern ssize_t iter_file_splice_write(struct pipe_inode_info *,
2892                struct file *, loff_t *, size_t, unsigned int);
2893extern ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe,
2894                struct file *out, loff_t *, size_t len, unsigned int flags);
2895extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
2896                loff_t *opos, size_t len, unsigned int flags);
2897
2898
2899extern void
2900file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping);
2901extern loff_t noop_llseek(struct file *file, loff_t offset, int whence);
2902extern loff_t no_llseek(struct file *file, loff_t offset, int whence);
2903extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize);
2904extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence);
2905extern loff_t generic_file_llseek_size(struct file *file, loff_t offset,
2906                int whence, loff_t maxsize, loff_t eof);
2907extern loff_t fixed_size_llseek(struct file *file, loff_t offset,
2908                int whence, loff_t size);
2909extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t);
2910extern loff_t no_seek_end_llseek(struct file *, loff_t, int);
2911extern int generic_file_open(struct inode * inode, struct file * filp);
2912extern int nonseekable_open(struct inode * inode, struct file * filp);
2913
2914#ifdef CONFIG_BLOCK
2915typedef void (dio_submit_t)(struct bio *bio, struct inode *inode,
2916                            loff_t file_offset);
2917
2918enum {
2919        /* need locking between buffered and direct access */
2920        DIO_LOCKING     = 0x01,
2921
2922        /* filesystem does not support filling holes */
2923        DIO_SKIP_HOLES  = 0x02,
2924
2925        /* filesystem can handle aio writes beyond i_size */
2926        DIO_ASYNC_EXTEND = 0x04,
2927
2928        /* inode/fs/bdev does not need truncate protection */
2929        DIO_SKIP_DIO_COUNT = 0x08,
2930};
2931
2932void dio_end_io(struct bio *bio);
2933
2934ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
2935                             struct block_device *bdev, struct iov_iter *iter,
2936                             get_block_t get_block,
2937                             dio_iodone_t end_io, dio_submit_t submit_io,
2938                             int flags);
2939
2940static inline ssize_t blockdev_direct_IO(struct kiocb *iocb,
2941                                         struct inode *inode,
2942                                         struct iov_iter *iter,
2943                                         get_block_t get_block)
2944{
2945        return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter,
2946                        get_block, NULL, NULL, DIO_LOCKING | DIO_SKIP_HOLES);
2947}
2948#endif
2949
2950void inode_dio_wait(struct inode *inode);
2951
2952/*
2953 * inode_dio_begin - signal start of a direct I/O requests
2954 * @inode: inode the direct I/O happens on
2955 *
2956 * This is called once we've finished processing a direct I/O request,
2957 * and is used to wake up callers waiting for direct I/O to be quiesced.
2958 */
2959static inline void inode_dio_begin(struct inode *inode)
2960{
2961        atomic_inc(&inode->i_dio_count);
2962}
2963
2964/*
2965 * inode_dio_end - signal finish of a direct I/O requests
2966 * @inode: inode the direct I/O happens on
2967 *
2968 * This is called once we've finished processing a direct I/O request,
2969 * and is used to wake up callers waiting for direct I/O to be quiesced.
2970 */
2971static inline void inode_dio_end(struct inode *inode)
2972{
2973        if (atomic_dec_and_test(&inode->i_dio_count))
2974                wake_up_bit(&inode->i_state, __I_DIO_WAKEUP);
2975}
2976
2977extern void inode_set_flags(struct inode *inode, unsigned int flags,
2978                            unsigned int mask);
2979
2980extern const struct file_operations generic_ro_fops;
2981
2982#define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m))
2983
2984extern int readlink_copy(char __user *, int, const char *);
2985extern int page_readlink(struct dentry *, char __user *, int);
2986extern const char *page_get_link(struct dentry *, struct inode *,
2987                                 struct delayed_call *);
2988extern void page_put_link(void *);
2989extern int __page_symlink(struct inode *inode, const char *symname, int len,
2990                int nofs);
2991extern int page_symlink(struct inode *inode, const char *symname, int len);
2992extern const struct inode_operations page_symlink_inode_operations;
2993extern void kfree_link(void *);
2994extern void generic_fillattr(struct inode *, struct kstat *);
2995extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int);
2996extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int);
2997void __inode_add_bytes(struct inode *inode, loff_t bytes);
2998void inode_add_bytes(struct inode *inode, loff_t bytes);
2999void __inode_sub_bytes(struct inode *inode, loff_t bytes);
3000void inode_sub_bytes(struct inode *inode, loff_t bytes);
3001loff_t inode_get_bytes(struct inode *inode);
3002void inode_set_bytes(struct inode *inode, loff_t bytes);
3003const char *simple_get_link(struct dentry *, struct inode *,
3004                            struct delayed_call *);
3005extern const struct inode_operations simple_symlink_inode_operations;
3006
3007extern int iterate_dir(struct file *, struct dir_context *);
3008
3009extern int vfs_statx(int, const char __user *, int, struct kstat *, u32);
3010extern int vfs_statx_fd(unsigned int, struct kstat *, u32, unsigned int);
3011
3012static inline int vfs_stat(const char __user *filename, struct kstat *stat)
3013{
3014        return vfs_statx(AT_FDCWD, filename, AT_NO_AUTOMOUNT,
3015                         stat, STATX_BASIC_STATS);
3016}
3017static inline int vfs_lstat(const char __user *name, struct kstat *stat)
3018{
3019        return vfs_statx(AT_FDCWD, name, AT_SYMLINK_NOFOLLOW | AT_NO_AUTOMOUNT,
3020                         stat, STATX_BASIC_STATS);
3021}
3022static inline int vfs_fstatat(int dfd, const char __user *filename,
3023                              struct kstat *stat, int flags)
3024{
3025        return vfs_statx(dfd, filename, flags | AT_NO_AUTOMOUNT,
3026                         stat, STATX_BASIC_STATS);
3027}
3028static inline int vfs_fstat(int fd, struct kstat *stat)
3029{
3030        return vfs_statx_fd(fd, stat, STATX_BASIC_STATS, 0);
3031}
3032
3033
3034extern const char *vfs_get_link(struct dentry *, struct delayed_call *);
3035extern int vfs_readlink(struct dentry *, char __user *, int);
3036
3037extern int __generic_block_fiemap(struct inode *inode,
3038                                  struct fiemap_extent_info *fieinfo,
3039                                  loff_t start, loff_t len,
3040                                  get_block_t *get_block);
3041extern int generic_block_fiemap(struct inode *inode,
3042                                struct fiemap_extent_info *fieinfo, u64 start,
3043                                u64 len, get_block_t *get_block);
3044
3045extern struct file_system_type *get_filesystem(struct file_system_type *fs);
3046extern void put_filesystem(struct file_system_type *fs);
3047extern struct file_system_type *get_fs_type(const char *name);
3048extern struct super_block *get_super(struct block_device *);
3049extern struct super_block *get_super_thawed(struct block_device *);
3050extern struct super_block *get_super_exclusive_thawed(struct block_device *bdev);
3051extern struct super_block *get_active_super(struct block_device *bdev);
3052extern void drop_super(struct super_block *sb);
3053extern void drop_super_exclusive(struct super_block *sb);
3054extern void iterate_supers(void (*)(struct super_block *, void *), void *);
3055extern void iterate_supers_type(struct file_system_type *,
3056                                void (*)(struct super_block *, void *), void *);
3057
3058extern int dcache_dir_open(struct inode *, struct file *);
3059extern int dcache_dir_close(struct inode *, struct file *);
3060extern loff_t dcache_dir_lseek(struct file *, loff_t, int);
3061extern int dcache_readdir(struct file *, struct dir_context *);
3062extern int simple_setattr(struct dentry *, struct iattr *);
3063extern int simple_getattr(const struct path *, struct kstat *, u32, unsigned int);
3064extern int simple_statfs(struct dentry *, struct kstatfs *);
3065extern int simple_open(struct inode *inode, struct file *file);
3066extern int simple_link(struct dentry *, struct inode *, struct dentry *);
3067extern int simple_unlink(struct inode *, struct dentry *);
3068extern int simple_rmdir(struct inode *, struct dentry *);
3069extern int simple_rename(struct inode *, struct dentry *,
3070                         struct inode *, struct dentry *, unsigned int);
3071extern int noop_fsync(struct file *, loff_t, loff_t, int);
3072extern int simple_empty(struct dentry *);
3073extern int simple_readpage(struct file *file, struct page *page);
3074extern int simple_write_begin(struct file *file, struct address_space *mapping,
3075                        loff_t pos, unsigned len, unsigned flags,
3076                        struct page **pagep, void **fsdata);
3077extern int simple_write_end(struct file *file, struct address_space *mapping,
3078                        loff_t pos, unsigned len, unsigned copied,
3079                        struct page *page, void *fsdata);
3080extern int always_delete_dentry(const struct dentry *);
3081extern struct inode *alloc_anon_inode(struct super_block *);
3082extern int simple_nosetlease(struct file *, long, struct file_lock **, void **);
3083extern const struct dentry_operations simple_dentry_operations;
3084
3085extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags);
3086extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *);
3087extern const struct file_operations simple_dir_operations;
3088extern const struct inode_operations simple_dir_inode_operations;
3089extern void make_empty_dir_inode(struct inode *inode);
3090extern bool is_empty_dir_inode(struct inode *inode);
3091struct tree_descr { const char *name; const struct file_operations *ops; int mode; };
3092struct dentry *d_alloc_name(struct dentry *, const char *);
3093extern int simple_fill_super(struct super_block *, unsigned long,
3094                             const struct tree_descr *);
3095extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count);
3096extern void simple_release_fs(struct vfsmount **mount, int *count);
3097
3098extern ssize_t simple_read_from_buffer(void __user *to, size_t count,
3099                        loff_t *ppos, const void *from, size_t available);
3100extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
3101                const void __user *from, size_t count);
3102
3103extern int __generic_file_fsync(struct file *, loff_t, loff_t, int);
3104extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
3105
3106extern int generic_check_addressable(unsigned, u64);
3107
3108#ifdef CONFIG_MIGRATION
3109extern int buffer_migrate_page(struct address_space *,
3110                                struct page *, struct page *,
3111                                enum migrate_mode);
3112#else
3113#define buffer_migrate_page NULL
3114#endif
3115
3116extern int setattr_prepare(struct dentry *, struct iattr *);
3117extern int inode_newsize_ok(const struct inode *, loff_t offset);
3118extern void setattr_copy(struct inode *inode, const struct iattr *attr);
3119
3120extern int file_update_time(struct file *file);
3121
3122static inline bool io_is_direct(struct file *filp)
3123{
3124        return (filp->f_flags & O_DIRECT) || IS_DAX(filp->f_mapping->host);
3125}
3126
3127static inline bool vma_is_dax(struct vm_area_struct *vma)
3128{
3129        return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host);
3130}
3131
3132static inline int iocb_flags(struct file *file)
3133{
3134        int res = 0;
3135        if (file->f_flags & O_APPEND)
3136                res |= IOCB_APPEND;
3137        if (io_is_direct(file))
3138                res |= IOCB_DIRECT;
3139        if ((file->f_flags & O_DSYNC) || IS_SYNC(file->f_mapping->host))
3140                res |= IOCB_DSYNC;
3141        if (file->f_flags & __O_SYNC)
3142                res |= IOCB_SYNC;
3143        return res;
3144}
3145
3146static inline int kiocb_set_rw_flags(struct kiocb *ki, int flags)
3147{
3148        if (unlikely(flags & ~RWF_SUPPORTED))
3149                return -EOPNOTSUPP;
3150
3151        if (flags & RWF_NOWAIT) {
3152                if (!(ki->ki_filp->f_mode & FMODE_AIO_NOWAIT))
3153                        return -EOPNOTSUPP;
3154                ki->ki_flags |= IOCB_NOWAIT;
3155        }
3156        if (flags & RWF_HIPRI)
3157                ki->ki_flags |= IOCB_HIPRI;
3158        if (flags & RWF_DSYNC)
3159                ki->ki_flags |= IOCB_DSYNC;
3160        if (flags & RWF_SYNC)
3161                ki->ki_flags |= (IOCB_DSYNC | IOCB_SYNC);
3162        return 0;
3163}
3164
3165static inline ino_t parent_ino(struct dentry *dentry)
3166{
3167        ino_t res;
3168
3169        /*
3170         * Don't strictly need d_lock here? If the parent ino could change
3171         * then surely we'd have a deeper race in the caller?
3172         */
3173        spin_lock(&dentry->d_lock);
3174        res = dentry->d_parent->d_inode->i_ino;
3175        spin_unlock(&dentry->d_lock);
3176        return res;
3177}
3178
3179/* Transaction based IO helpers */
3180
3181/*
3182 * An argresp is stored in an allocated page and holds the
3183 * size of the argument or response, along with its content
3184 */
3185struct simple_transaction_argresp {
3186        ssize_t size;
3187        char data[0];
3188};
3189
3190#define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp))
3191
3192char *simple_transaction_get(struct file *file, const char __user *buf,
3193                                size_t size);
3194ssize_t simple_transaction_read(struct file *file, char __user *buf,
3195                                size_t size, loff_t *pos);
3196int simple_transaction_release(struct inode *inode, struct file *file);
3197
3198void simple_transaction_set(struct file *file, size_t n);
3199
3200/*
3201 * simple attribute files
3202 *
3203 * These attributes behave similar to those in sysfs:
3204 *
3205 * Writing to an attribute immediately sets a value, an open file can be
3206 * written to multiple times.
3207 *
3208 * Reading from an attribute creates a buffer from the value that might get
3209 * read with multiple read calls. When the attribute has been read
3210 * completely, no further read calls are possible until the file is opened
3211 * again.
3212 *
3213 * All attributes contain a text representation of a numeric value
3214 * that are accessed with the get() and set() functions.
3215 */
3216#define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt)            \
3217static int __fops ## _open(struct inode *inode, struct file *file)      \
3218{                                                                       \
3219        __simple_attr_check_format(__fmt, 0ull);                        \
3220        return simple_attr_open(inode, file, __get, __set, __fmt);      \
3221}                                                                       \
3222static const struct file_operations __fops = {                          \
3223        .owner   = THIS_MODULE,                                         \
3224        .open    = __fops ## _open,                                     \
3225        .release = simple_attr_release,                                 \
3226        .read    = simple_attr_read,                                    \
3227        .write   = simple_attr_write,                                   \
3228        .llseek  = generic_file_llseek,                                 \
3229}
3230
3231static inline __printf(1, 2)
3232void __simple_attr_check_format(const char *fmt, ...)
3233{
3234        /* don't do anything, just let the compiler check the arguments; */
3235}
3236
3237int simple_attr_open(struct inode *inode, struct file *file,
3238                     int (*get)(void *, u64 *), int (*set)(void *, u64),
3239                     const char *fmt);
3240int simple_attr_release(struct inode *inode, struct file *file);
3241ssize_t simple_attr_read(struct file *file, char __user *buf,
3242                         size_t len, loff_t *ppos);
3243ssize_t simple_attr_write(struct file *file, const char __user *buf,
3244                          size_t len, loff_t *ppos);
3245
3246struct ctl_table;
3247int proc_nr_files(struct ctl_table *table, int write,
3248                  void __user *buffer, size_t *lenp, loff_t *ppos);
3249int proc_nr_dentry(struct ctl_table *table, int write,
3250                  void __user *buffer, size_t *lenp, loff_t *ppos);
3251int proc_nr_inodes(struct ctl_table *table, int write,
3252                   void __user *buffer, size_t *lenp, loff_t *ppos);
3253int __init get_filesystem_list(char *buf);
3254
3255#define __FMODE_EXEC            ((__force int) FMODE_EXEC)
3256#define __FMODE_NONOTIFY        ((__force int) FMODE_NONOTIFY)
3257
3258#define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
3259#define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \
3260                                            (flag & __FMODE_NONOTIFY)))
3261
3262static inline bool is_sxid(umode_t mode)
3263{
3264        return (mode & S_ISUID) || ((mode & S_ISGID) && (mode & S_IXGRP));
3265}
3266
3267static inline int check_sticky(struct inode *dir, struct inode *inode)
3268{
3269        if (!(dir->i_mode & S_ISVTX))
3270                return 0;
3271
3272        return __check_sticky(dir, inode);
3273}
3274
3275static inline void inode_has_no_xattr(struct inode *inode)
3276{
3277        if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & MS_NOSEC))
3278                inode->i_flags |= S_NOSEC;
3279}
3280
3281static inline bool is_root_inode(struct inode *inode)
3282{
3283        return inode == inode->i_sb->s_root->d_inode;
3284}
3285
3286static inline bool dir_emit(struct dir_context *ctx,
3287                            const char *name, int namelen,
3288                            u64 ino, unsigned type)
3289{
3290        return ctx->actor(ctx, name, namelen, ctx->pos, ino, type) == 0;
3291}
3292static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx)
3293{
3294        return ctx->actor(ctx, ".", 1, ctx->pos,
3295                          file->f_path.dentry->d_inode->i_ino, DT_DIR) == 0;
3296}
3297static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx)
3298{
3299        return ctx->actor(ctx, "..", 2, ctx->pos,
3300                          parent_ino(file->f_path.dentry), DT_DIR) == 0;
3301}
3302static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx)
3303{
3304        if (ctx->pos == 0) {
3305                if (!dir_emit_dot(file, ctx))
3306                        return false;
3307                ctx->pos = 1;
3308        }
3309        if (ctx->pos == 1) {
3310                if (!dir_emit_dotdot(file, ctx))
3311                        return false;
3312                ctx->pos = 2;
3313        }
3314        return true;
3315}
3316static inline bool dir_relax(struct inode *inode)
3317{
3318        inode_unlock(inode);
3319        inode_lock(inode);
3320        return !IS_DEADDIR(inode);
3321}
3322
3323static inline bool dir_relax_shared(struct inode *inode)
3324{
3325        inode_unlock_shared(inode);
3326        inode_lock_shared(inode);
3327        return !IS_DEADDIR(inode);
3328}
3329
3330extern bool path_noexec(const struct path *path);
3331extern void inode_nohighmem(struct inode *inode);
3332
3333#endif /* _LINUX_FS_H */
3334