linux/Documentation/filesystems/vfs.rst
<<
>>
Prefs
   1.. SPDX-License-Identifier: GPL-2.0
   2
   3=========================================
   4Overview of the Linux Virtual File System
   5=========================================
   6
   7Original author: Richard Gooch <rgooch@atnf.csiro.au>
   8
   9- Copyright (C) 1999 Richard Gooch
  10- Copyright (C) 2005 Pekka Enberg
  11
  12
  13Introduction
  14============
  15
  16The Virtual File System (also known as the Virtual Filesystem Switch) is
  17the software layer in the kernel that provides the filesystem interface
  18to userspace programs.  It also provides an abstraction within the
  19kernel which allows different filesystem implementations to coexist.
  20
  21VFS system calls open(2), stat(2), read(2), write(2), chmod(2) and so on
  22are called from a process context.  Filesystem locking is described in
  23the document Documentation/filesystems/locking.rst.
  24
  25
  26Directory Entry Cache (dcache)
  27------------------------------
  28
  29The VFS implements the open(2), stat(2), chmod(2), and similar system
  30calls.  The pathname argument that is passed to them is used by the VFS
  31to search through the directory entry cache (also known as the dentry
  32cache or dcache).  This provides a very fast look-up mechanism to
  33translate a pathname (filename) into a specific dentry.  Dentries live
  34in RAM and are never saved to disc: they exist only for performance.
  35
  36The dentry cache is meant to be a view into your entire filespace.  As
  37most computers cannot fit all dentries in the RAM at the same time, some
  38bits of the cache are missing.  In order to resolve your pathname into a
  39dentry, the VFS may have to resort to creating dentries along the way,
  40and then loading the inode.  This is done by looking up the inode.
  41
  42
  43The Inode Object
  44----------------
  45
  46An individual dentry usually has a pointer to an inode.  Inodes are
  47filesystem objects such as regular files, directories, FIFOs and other
  48beasts.  They live either on the disc (for block device filesystems) or
  49in the memory (for pseudo filesystems).  Inodes that live on the disc
  50are copied into the memory when required and changes to the inode are
  51written back to disc.  A single inode can be pointed to by multiple
  52dentries (hard links, for example, do this).
  53
  54To look up an inode requires that the VFS calls the lookup() method of
  55the parent directory inode.  This method is installed by the specific
  56filesystem implementation that the inode lives in.  Once the VFS has the
  57required dentry (and hence the inode), we can do all those boring things
  58like open(2) the file, or stat(2) it to peek at the inode data.  The
  59stat(2) operation is fairly simple: once the VFS has the dentry, it
  60peeks at the inode data and passes some of it back to userspace.
  61
  62
  63The File Object
  64---------------
  65
  66Opening a file requires another operation: allocation of a file
  67structure (this is the kernel-side implementation of file descriptors).
  68The freshly allocated file structure is initialized with a pointer to
  69the dentry and a set of file operation member functions.  These are
  70taken from the inode data.  The open() file method is then called so the
  71specific filesystem implementation can do its work.  You can see that
  72this is another switch performed by the VFS.  The file structure is
  73placed into the file descriptor table for the process.
  74
  75Reading, writing and closing files (and other assorted VFS operations)
  76is done by using the userspace file descriptor to grab the appropriate
  77file structure, and then calling the required file structure method to
  78do whatever is required.  For as long as the file is open, it keeps the
  79dentry in use, which in turn means that the VFS inode is still in use.
  80
  81
  82Registering and Mounting a Filesystem
  83=====================================
  84
  85To register and unregister a filesystem, use the following API
  86functions:
  87
  88.. code-block:: c
  89
  90        #include <linux/fs.h>
  91
  92        extern int register_filesystem(struct file_system_type *);
  93        extern int unregister_filesystem(struct file_system_type *);
  94
  95The passed struct file_system_type describes your filesystem.  When a
  96request is made to mount a filesystem onto a directory in your
  97namespace, the VFS will call the appropriate mount() method for the
  98specific filesystem.  New vfsmount referring to the tree returned by
  99->mount() will be attached to the mountpoint, so that when pathname
 100resolution reaches the mountpoint it will jump into the root of that
 101vfsmount.
 102
 103You can see all filesystems that are registered to the kernel in the
 104file /proc/filesystems.
 105
 106
 107struct file_system_type
 108-----------------------
 109
 110This describes the filesystem.  As of kernel 2.6.39, the following
 111members are defined:
 112
 113.. code-block:: c
 114
 115        struct file_system_type {
 116                const char *name;
 117                int fs_flags;
 118                struct dentry *(*mount) (struct file_system_type *, int,
 119                                         const char *, void *);
 120                void (*kill_sb) (struct super_block *);
 121                struct module *owner;
 122                struct file_system_type * next;
 123                struct list_head fs_supers;
 124                struct lock_class_key s_lock_key;
 125                struct lock_class_key s_umount_key;
 126        };
 127
 128``name``
 129        the name of the filesystem type, such as "ext2", "iso9660",
 130        "msdos" and so on
 131
 132``fs_flags``
 133        various flags (i.e. FS_REQUIRES_DEV, FS_NO_DCACHE, etc.)
 134
 135``mount``
 136        the method to call when a new instance of this filesystem should
 137        be mounted
 138
 139``kill_sb``
 140        the method to call when an instance of this filesystem should be
 141        shut down
 142
 143
 144``owner``
 145        for internal VFS use: you should initialize this to THIS_MODULE
 146        in most cases.
 147
 148``next``
 149        for internal VFS use: you should initialize this to NULL
 150
 151  s_lock_key, s_umount_key: lockdep-specific
 152
 153The mount() method has the following arguments:
 154
 155``struct file_system_type *fs_type``
 156        describes the filesystem, partly initialized by the specific
 157        filesystem code
 158
 159``int flags``
 160        mount flags
 161
 162``const char *dev_name``
 163        the device name we are mounting.
 164
 165``void *data``
 166        arbitrary mount options, usually comes as an ASCII string (see
 167        "Mount Options" section)
 168
 169The mount() method must return the root dentry of the tree requested by
 170caller.  An active reference to its superblock must be grabbed and the
 171superblock must be locked.  On failure it should return ERR_PTR(error).
 172
 173The arguments match those of mount(2) and their interpretation depends
 174on filesystem type.  E.g. for block filesystems, dev_name is interpreted
 175as block device name, that device is opened and if it contains a
 176suitable filesystem image the method creates and initializes struct
 177super_block accordingly, returning its root dentry to caller.
 178
 179->mount() may choose to return a subtree of existing filesystem - it
 180doesn't have to create a new one.  The main result from the caller's
 181point of view is a reference to dentry at the root of (sub)tree to be
 182attached; creation of new superblock is a common side effect.
 183
 184The most interesting member of the superblock structure that the mount()
 185method fills in is the "s_op" field.  This is a pointer to a "struct
 186super_operations" which describes the next level of the filesystem
 187implementation.
 188
 189Usually, a filesystem uses one of the generic mount() implementations
 190and provides a fill_super() callback instead.  The generic variants are:
 191
 192``mount_bdev``
 193        mount a filesystem residing on a block device
 194
 195``mount_nodev``
 196        mount a filesystem that is not backed by a device
 197
 198``mount_single``
 199        mount a filesystem which shares the instance between all mounts
 200
 201A fill_super() callback implementation has the following arguments:
 202
 203``struct super_block *sb``
 204        the superblock structure.  The callback must initialize this
 205        properly.
 206
 207``void *data``
 208        arbitrary mount options, usually comes as an ASCII string (see
 209        "Mount Options" section)
 210
 211``int silent``
 212        whether or not to be silent on error
 213
 214
 215The Superblock Object
 216=====================
 217
 218A superblock object represents a mounted filesystem.
 219
 220
 221struct super_operations
 222-----------------------
 223
 224This describes how the VFS can manipulate the superblock of your
 225filesystem.  As of kernel 2.6.22, the following members are defined:
 226
 227.. code-block:: c
 228
 229        struct super_operations {
 230                struct inode *(*alloc_inode)(struct super_block *sb);
 231                void (*destroy_inode)(struct inode *);
 232
 233                void (*dirty_inode) (struct inode *, int flags);
 234                int (*write_inode) (struct inode *, int);
 235                void (*drop_inode) (struct inode *);
 236                void (*delete_inode) (struct inode *);
 237                void (*put_super) (struct super_block *);
 238                int (*sync_fs)(struct super_block *sb, int wait);
 239                int (*freeze_fs) (struct super_block *);
 240                int (*unfreeze_fs) (struct super_block *);
 241                int (*statfs) (struct dentry *, struct kstatfs *);
 242                int (*remount_fs) (struct super_block *, int *, char *);
 243                void (*clear_inode) (struct inode *);
 244                void (*umount_begin) (struct super_block *);
 245
 246                int (*show_options)(struct seq_file *, struct dentry *);
 247
 248                ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
 249                ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
 250                int (*nr_cached_objects)(struct super_block *);
 251                void (*free_cached_objects)(struct super_block *, int);
 252        };
 253
 254All methods are called without any locks being held, unless otherwise
 255noted.  This means that most methods can block safely.  All methods are
 256only called from a process context (i.e. not from an interrupt handler
 257or bottom half).
 258
 259``alloc_inode``
 260        this method is called by alloc_inode() to allocate memory for
 261        struct inode and initialize it.  If this function is not
 262        defined, a simple 'struct inode' is allocated.  Normally
 263        alloc_inode will be used to allocate a larger structure which
 264        contains a 'struct inode' embedded within it.
 265
 266``destroy_inode``
 267        this method is called by destroy_inode() to release resources
 268        allocated for struct inode.  It is only required if
 269        ->alloc_inode was defined and simply undoes anything done by
 270        ->alloc_inode.
 271
 272``dirty_inode``
 273        this method is called by the VFS when an inode is marked dirty.
 274        This is specifically for the inode itself being marked dirty,
 275        not its data.  If the update needs to be persisted by fdatasync(),
 276        then I_DIRTY_DATASYNC will be set in the flags argument.
 277
 278``write_inode``
 279        this method is called when the VFS needs to write an inode to
 280        disc.  The second parameter indicates whether the write should
 281        be synchronous or not, not all filesystems check this flag.
 282
 283``drop_inode``
 284        called when the last access to the inode is dropped, with the
 285        inode->i_lock spinlock held.
 286
 287        This method should be either NULL (normal UNIX filesystem
 288        semantics) or "generic_delete_inode" (for filesystems that do
 289        not want to cache inodes - causing "delete_inode" to always be
 290        called regardless of the value of i_nlink)
 291
 292        The "generic_delete_inode()" behavior is equivalent to the old
 293        practice of using "force_delete" in the put_inode() case, but
 294        does not have the races that the "force_delete()" approach had.
 295
 296``delete_inode``
 297        called when the VFS wants to delete an inode
 298
 299``put_super``
 300        called when the VFS wishes to free the superblock
 301        (i.e. unmount).  This is called with the superblock lock held
 302
 303``sync_fs``
 304        called when VFS is writing out all dirty data associated with a
 305        superblock.  The second parameter indicates whether the method
 306        should wait until the write out has been completed.  Optional.
 307
 308``freeze_fs``
 309        called when VFS is locking a filesystem and forcing it into a
 310        consistent state.  This method is currently used by the Logical
 311        Volume Manager (LVM).
 312
 313``unfreeze_fs``
 314        called when VFS is unlocking a filesystem and making it writable
 315        again.
 316
 317``statfs``
 318        called when the VFS needs to get filesystem statistics.
 319
 320``remount_fs``
 321        called when the filesystem is remounted.  This is called with
 322        the kernel lock held
 323
 324``clear_inode``
 325        called then the VFS clears the inode.  Optional
 326
 327``umount_begin``
 328        called when the VFS is unmounting a filesystem.
 329
 330``show_options``
 331        called by the VFS to show mount options for /proc/<pid>/mounts.
 332        (see "Mount Options" section)
 333
 334``quota_read``
 335        called by the VFS to read from filesystem quota file.
 336
 337``quota_write``
 338        called by the VFS to write to filesystem quota file.
 339
 340``nr_cached_objects``
 341        called by the sb cache shrinking function for the filesystem to
 342        return the number of freeable cached objects it contains.
 343        Optional.
 344
 345``free_cache_objects``
 346        called by the sb cache shrinking function for the filesystem to
 347        scan the number of objects indicated to try to free them.
 348        Optional, but any filesystem implementing this method needs to
 349        also implement ->nr_cached_objects for it to be called
 350        correctly.
 351
 352        We can't do anything with any errors that the filesystem might
 353        encountered, hence the void return type.  This will never be
 354        called if the VM is trying to reclaim under GFP_NOFS conditions,
 355        hence this method does not need to handle that situation itself.
 356
 357        Implementations must include conditional reschedule calls inside
 358        any scanning loop that is done.  This allows the VFS to
 359        determine appropriate scan batch sizes without having to worry
 360        about whether implementations will cause holdoff problems due to
 361        large scan batch sizes.
 362
 363Whoever sets up the inode is responsible for filling in the "i_op"
 364field.  This is a pointer to a "struct inode_operations" which describes
 365the methods that can be performed on individual inodes.
 366
 367
 368struct xattr_handlers
 369---------------------
 370
 371On filesystems that support extended attributes (xattrs), the s_xattr
 372superblock field points to a NULL-terminated array of xattr handlers.
 373Extended attributes are name:value pairs.
 374
 375``name``
 376        Indicates that the handler matches attributes with the specified
 377        name (such as "system.posix_acl_access"); the prefix field must
 378        be NULL.
 379
 380``prefix``
 381        Indicates that the handler matches all attributes with the
 382        specified name prefix (such as "user."); the name field must be
 383        NULL.
 384
 385``list``
 386        Determine if attributes matching this xattr handler should be
 387        listed for a particular dentry.  Used by some listxattr
 388        implementations like generic_listxattr.
 389
 390``get``
 391        Called by the VFS to get the value of a particular extended
 392        attribute.  This method is called by the getxattr(2) system
 393        call.
 394
 395``set``
 396        Called by the VFS to set the value of a particular extended
 397        attribute.  When the new value is NULL, called to remove a
 398        particular extended attribute.  This method is called by the
 399        setxattr(2) and removexattr(2) system calls.
 400
 401When none of the xattr handlers of a filesystem match the specified
 402attribute name or when a filesystem doesn't support extended attributes,
 403the various ``*xattr(2)`` system calls return -EOPNOTSUPP.
 404
 405
 406The Inode Object
 407================
 408
 409An inode object represents an object within the filesystem.
 410
 411
 412struct inode_operations
 413-----------------------
 414
 415This describes how the VFS can manipulate an inode in your filesystem.
 416As of kernel 2.6.22, the following members are defined:
 417
 418.. code-block:: c
 419
 420        struct inode_operations {
 421                int (*create) (struct user_namespace *, struct inode *,struct dentry *, umode_t, bool);
 422                struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
 423                int (*link) (struct dentry *,struct inode *,struct dentry *);
 424                int (*unlink) (struct inode *,struct dentry *);
 425                int (*symlink) (struct user_namespace *, struct inode *,struct dentry *,const char *);
 426                int (*mkdir) (struct user_namespace *, struct inode *,struct dentry *,umode_t);
 427                int (*rmdir) (struct inode *,struct dentry *);
 428                int (*mknod) (struct user_namespace *, struct inode *,struct dentry *,umode_t,dev_t);
 429                int (*rename) (struct user_namespace *, struct inode *, struct dentry *,
 430                               struct inode *, struct dentry *, unsigned int);
 431                int (*readlink) (struct dentry *, char __user *,int);
 432                const char *(*get_link) (struct dentry *, struct inode *,
 433                                         struct delayed_call *);
 434                int (*permission) (struct user_namespace *, struct inode *, int);
 435                struct posix_acl * (*get_acl)(struct inode *, int, bool);
 436                int (*setattr) (struct user_namespace *, struct dentry *, struct iattr *);
 437                int (*getattr) (struct user_namespace *, const struct path *, struct kstat *, u32, unsigned int);
 438                ssize_t (*listxattr) (struct dentry *, char *, size_t);
 439                void (*update_time)(struct inode *, struct timespec *, int);
 440                int (*atomic_open)(struct inode *, struct dentry *, struct file *,
 441                                   unsigned open_flag, umode_t create_mode);
 442                int (*tmpfile) (struct user_namespace *, struct inode *, struct dentry *, umode_t);
 443                int (*set_acl)(struct user_namespace *, struct inode *, struct posix_acl *, int);
 444                int (*fileattr_set)(struct user_namespace *mnt_userns,
 445                                    struct dentry *dentry, struct fileattr *fa);
 446                int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa);
 447        };
 448
 449Again, all methods are called without any locks being held, unless
 450otherwise noted.
 451
 452``create``
 453        called by the open(2) and creat(2) system calls.  Only required
 454        if you want to support regular files.  The dentry you get should
 455        not have an inode (i.e. it should be a negative dentry).  Here
 456        you will probably call d_instantiate() with the dentry and the
 457        newly created inode
 458
 459``lookup``
 460        called when the VFS needs to look up an inode in a parent
 461        directory.  The name to look for is found in the dentry.  This
 462        method must call d_add() to insert the found inode into the
 463        dentry.  The "i_count" field in the inode structure should be
 464        incremented.  If the named inode does not exist a NULL inode
 465        should be inserted into the dentry (this is called a negative
 466        dentry).  Returning an error code from this routine must only be
 467        done on a real error, otherwise creating inodes with system
 468        calls like create(2), mknod(2), mkdir(2) and so on will fail.
 469        If you wish to overload the dentry methods then you should
 470        initialise the "d_dop" field in the dentry; this is a pointer to
 471        a struct "dentry_operations".  This method is called with the
 472        directory inode semaphore held
 473
 474``link``
 475        called by the link(2) system call.  Only required if you want to
 476        support hard links.  You will probably need to call
 477        d_instantiate() just as you would in the create() method
 478
 479``unlink``
 480        called by the unlink(2) system call.  Only required if you want
 481        to support deleting inodes
 482
 483``symlink``
 484        called by the symlink(2) system call.  Only required if you want
 485        to support symlinks.  You will probably need to call
 486        d_instantiate() just as you would in the create() method
 487
 488``mkdir``
 489        called by the mkdir(2) system call.  Only required if you want
 490        to support creating subdirectories.  You will probably need to
 491        call d_instantiate() just as you would in the create() method
 492
 493``rmdir``
 494        called by the rmdir(2) system call.  Only required if you want
 495        to support deleting subdirectories
 496
 497``mknod``
 498        called by the mknod(2) system call to create a device (char,
 499        block) inode or a named pipe (FIFO) or socket.  Only required if
 500        you want to support creating these types of inodes.  You will
 501        probably need to call d_instantiate() just as you would in the
 502        create() method
 503
 504``rename``
 505        called by the rename(2) system call to rename the object to have
 506        the parent and name given by the second inode and dentry.
 507
 508        The filesystem must return -EINVAL for any unsupported or
 509        unknown flags.  Currently the following flags are implemented:
 510        (1) RENAME_NOREPLACE: this flag indicates that if the target of
 511        the rename exists the rename should fail with -EEXIST instead of
 512        replacing the target.  The VFS already checks for existence, so
 513        for local filesystems the RENAME_NOREPLACE implementation is
 514        equivalent to plain rename.
 515        (2) RENAME_EXCHANGE: exchange source and target.  Both must
 516        exist; this is checked by the VFS.  Unlike plain rename, source
 517        and target may be of different type.
 518
 519``get_link``
 520        called by the VFS to follow a symbolic link to the inode it
 521        points to.  Only required if you want to support symbolic links.
 522        This method returns the symlink body to traverse (and possibly
 523        resets the current position with nd_jump_link()).  If the body
 524        won't go away until the inode is gone, nothing else is needed;
 525        if it needs to be otherwise pinned, arrange for its release by
 526        having get_link(..., ..., done) do set_delayed_call(done,
 527        destructor, argument).  In that case destructor(argument) will
 528        be called once VFS is done with the body you've returned.  May
 529        be called in RCU mode; that is indicated by NULL dentry
 530        argument.  If request can't be handled without leaving RCU mode,
 531        have it return ERR_PTR(-ECHILD).
 532
 533        If the filesystem stores the symlink target in ->i_link, the
 534        VFS may use it directly without calling ->get_link(); however,
 535        ->get_link() must still be provided.  ->i_link must not be
 536        freed until after an RCU grace period.  Writing to ->i_link
 537        post-iget() time requires a 'release' memory barrier.
 538
 539``readlink``
 540        this is now just an override for use by readlink(2) for the
 541        cases when ->get_link uses nd_jump_link() or object is not in
 542        fact a symlink.  Normally filesystems should only implement
 543        ->get_link for symlinks and readlink(2) will automatically use
 544        that.
 545
 546``permission``
 547        called by the VFS to check for access rights on a POSIX-like
 548        filesystem.
 549
 550        May be called in rcu-walk mode (mask & MAY_NOT_BLOCK).  If in
 551        rcu-walk mode, the filesystem must check the permission without
 552        blocking or storing to the inode.
 553
 554        If a situation is encountered that rcu-walk cannot handle,
 555        return
 556        -ECHILD and it will be called again in ref-walk mode.
 557
 558``setattr``
 559        called by the VFS to set attributes for a file.  This method is
 560        called by chmod(2) and related system calls.
 561
 562``getattr``
 563        called by the VFS to get attributes of a file.  This method is
 564        called by stat(2) and related system calls.
 565
 566``listxattr``
 567        called by the VFS to list all extended attributes for a given
 568        file.  This method is called by the listxattr(2) system call.
 569
 570``update_time``
 571        called by the VFS to update a specific time or the i_version of
 572        an inode.  If this is not defined the VFS will update the inode
 573        itself and call mark_inode_dirty_sync.
 574
 575``atomic_open``
 576        called on the last component of an open.  Using this optional
 577        method the filesystem can look up, possibly create and open the
 578        file in one atomic operation.  If it wants to leave actual
 579        opening to the caller (e.g. if the file turned out to be a
 580        symlink, device, or just something filesystem won't do atomic
 581        open for), it may signal this by returning finish_no_open(file,
 582        dentry).  This method is only called if the last component is
 583        negative or needs lookup.  Cached positive dentries are still
 584        handled by f_op->open().  If the file was created, FMODE_CREATED
 585        flag should be set in file->f_mode.  In case of O_EXCL the
 586        method must only succeed if the file didn't exist and hence
 587        FMODE_CREATED shall always be set on success.
 588
 589``tmpfile``
 590        called in the end of O_TMPFILE open().  Optional, equivalent to
 591        atomically creating, opening and unlinking a file in given
 592        directory.
 593
 594``fileattr_get``
 595        called on ioctl(FS_IOC_GETFLAGS) and ioctl(FS_IOC_FSGETXATTR) to
 596        retrieve miscellaneous file flags and attributes.  Also called
 597        before the relevant SET operation to check what is being changed
 598        (in this case with i_rwsem locked exclusive).  If unset, then
 599        fall back to f_op->ioctl().
 600
 601``fileattr_set``
 602        called on ioctl(FS_IOC_SETFLAGS) and ioctl(FS_IOC_FSSETXATTR) to
 603        change miscellaneous file flags and attributes.  Callers hold
 604        i_rwsem exclusive.  If unset, then fall back to f_op->ioctl().
 605
 606
 607The Address Space Object
 608========================
 609
 610The address space object is used to group and manage pages in the page
 611cache.  It can be used to keep track of the pages in a file (or anything
 612else) and also track the mapping of sections of the file into process
 613address spaces.
 614
 615There are a number of distinct yet related services that an
 616address-space can provide.  These include communicating memory pressure,
 617page lookup by address, and keeping track of pages tagged as Dirty or
 618Writeback.
 619
 620The first can be used independently to the others.  The VM can try to
 621either write dirty pages in order to clean them, or release clean pages
 622in order to reuse them.  To do this it can call the ->writepage method
 623on dirty pages, and ->releasepage on clean pages with PagePrivate set.
 624Clean pages without PagePrivate and with no external references will be
 625released without notice being given to the address_space.
 626
 627To achieve this functionality, pages need to be placed on an LRU with
 628lru_cache_add and mark_page_active needs to be called whenever the page
 629is used.
 630
 631Pages are normally kept in a radix tree index by ->index.  This tree
 632maintains information about the PG_Dirty and PG_Writeback status of each
 633page, so that pages with either of these flags can be found quickly.
 634
 635The Dirty tag is primarily used by mpage_writepages - the default
 636->writepages method.  It uses the tag to find dirty pages to call
 637->writepage on.  If mpage_writepages is not used (i.e. the address
 638provides its own ->writepages) , the PAGECACHE_TAG_DIRTY tag is almost
 639unused.  write_inode_now and sync_inode do use it (through
 640__sync_single_inode) to check if ->writepages has been successful in
 641writing out the whole address_space.
 642
 643The Writeback tag is used by filemap*wait* and sync_page* functions, via
 644filemap_fdatawait_range, to wait for all writeback to complete.
 645
 646An address_space handler may attach extra information to a page,
 647typically using the 'private' field in the 'struct page'.  If such
 648information is attached, the PG_Private flag should be set.  This will
 649cause various VM routines to make extra calls into the address_space
 650handler to deal with that data.
 651
 652An address space acts as an intermediate between storage and
 653application.  Data is read into the address space a whole page at a
 654time, and provided to the application either by copying of the page, or
 655by memory-mapping the page.  Data is written into the address space by
 656the application, and then written-back to storage typically in whole
 657pages, however the address_space has finer control of write sizes.
 658
 659The read process essentially only requires 'readpage'.  The write
 660process is more complicated and uses write_begin/write_end or
 661set_page_dirty to write data into the address_space, and writepage and
 662writepages to writeback data to storage.
 663
 664Adding and removing pages to/from an address_space is protected by the
 665inode's i_mutex.
 666
 667When data is written to a page, the PG_Dirty flag should be set.  It
 668typically remains set until writepage asks for it to be written.  This
 669should clear PG_Dirty and set PG_Writeback.  It can be actually written
 670at any point after PG_Dirty is clear.  Once it is known to be safe,
 671PG_Writeback is cleared.
 672
 673Writeback makes use of a writeback_control structure to direct the
 674operations.  This gives the writepage and writepages operations some
 675information about the nature of and reason for the writeback request,
 676and the constraints under which it is being done.  It is also used to
 677return information back to the caller about the result of a writepage or
 678writepages request.
 679
 680
 681Handling errors during writeback
 682--------------------------------
 683
 684Most applications that do buffered I/O will periodically call a file
 685synchronization call (fsync, fdatasync, msync or sync_file_range) to
 686ensure that data written has made it to the backing store.  When there
 687is an error during writeback, they expect that error to be reported when
 688a file sync request is made.  After an error has been reported on one
 689request, subsequent requests on the same file descriptor should return
 6900, unless further writeback errors have occurred since the previous file
 691syncronization.
 692
 693Ideally, the kernel would report errors only on file descriptions on
 694which writes were done that subsequently failed to be written back.  The
 695generic pagecache infrastructure does not track the file descriptions
 696that have dirtied each individual page however, so determining which
 697file descriptors should get back an error is not possible.
 698
 699Instead, the generic writeback error tracking infrastructure in the
 700kernel settles for reporting errors to fsync on all file descriptions
 701that were open at the time that the error occurred.  In a situation with
 702multiple writers, all of them will get back an error on a subsequent
 703fsync, even if all of the writes done through that particular file
 704descriptor succeeded (or even if there were no writes on that file
 705descriptor at all).
 706
 707Filesystems that wish to use this infrastructure should call
 708mapping_set_error to record the error in the address_space when it
 709occurs.  Then, after writing back data from the pagecache in their
 710file->fsync operation, they should call file_check_and_advance_wb_err to
 711ensure that the struct file's error cursor has advanced to the correct
 712point in the stream of errors emitted by the backing device(s).
 713
 714
 715struct address_space_operations
 716-------------------------------
 717
 718This describes how the VFS can manipulate mapping of a file to page
 719cache in your filesystem.  The following members are defined:
 720
 721.. code-block:: c
 722
 723        struct address_space_operations {
 724                int (*writepage)(struct page *page, struct writeback_control *wbc);
 725                int (*readpage)(struct file *, struct page *);
 726                int (*writepages)(struct address_space *, struct writeback_control *);
 727                int (*set_page_dirty)(struct page *page);
 728                void (*readahead)(struct readahead_control *);
 729                int (*readpages)(struct file *filp, struct address_space *mapping,
 730                                 struct list_head *pages, unsigned nr_pages);
 731                int (*write_begin)(struct file *, struct address_space *mapping,
 732                                   loff_t pos, unsigned len, unsigned flags,
 733                                struct page **pagep, void **fsdata);
 734                int (*write_end)(struct file *, struct address_space *mapping,
 735                                 loff_t pos, unsigned len, unsigned copied,
 736                                 struct page *page, void *fsdata);
 737                sector_t (*bmap)(struct address_space *, sector_t);
 738                void (*invalidatepage) (struct page *, unsigned int, unsigned int);
 739                int (*releasepage) (struct page *, int);
 740                void (*freepage)(struct page *);
 741                ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter);
 742                /* isolate a page for migration */
 743                bool (*isolate_page) (struct page *, isolate_mode_t);
 744                /* migrate the contents of a page to the specified target */
 745                int (*migratepage) (struct page *, struct page *);
 746                /* put migration-failed page back to right list */
 747                void (*putback_page) (struct page *);
 748                int (*launder_page) (struct page *);
 749
 750                int (*is_partially_uptodate) (struct page *, unsigned long,
 751                                              unsigned long);
 752                void (*is_dirty_writeback) (struct page *, bool *, bool *);
 753                int (*error_remove_page) (struct mapping *mapping, struct page *page);
 754                int (*swap_activate)(struct file *);
 755                int (*swap_deactivate)(struct file *);
 756        };
 757
 758``writepage``
 759        called by the VM to write a dirty page to backing store.  This
 760        may happen for data integrity reasons (i.e. 'sync'), or to free
 761        up memory (flush).  The difference can be seen in
 762        wbc->sync_mode.  The PG_Dirty flag has been cleared and
 763        PageLocked is true.  writepage should start writeout, should set
 764        PG_Writeback, and should make sure the page is unlocked, either
 765        synchronously or asynchronously when the write operation
 766        completes.
 767
 768        If wbc->sync_mode is WB_SYNC_NONE, ->writepage doesn't have to
 769        try too hard if there are problems, and may choose to write out
 770        other pages from the mapping if that is easier (e.g. due to
 771        internal dependencies).  If it chooses not to start writeout, it
 772        should return AOP_WRITEPAGE_ACTIVATE so that the VM will not
 773        keep calling ->writepage on that page.
 774
 775        See the file "Locking" for more details.
 776
 777``readpage``
 778        called by the VM to read a page from backing store.  The page
 779        will be Locked when readpage is called, and should be unlocked
 780        and marked uptodate once the read completes.  If ->readpage
 781        discovers that it needs to unlock the page for some reason, it
 782        can do so, and then return AOP_TRUNCATED_PAGE.  In this case,
 783        the page will be relocated, relocked and if that all succeeds,
 784        ->readpage will be called again.
 785
 786``writepages``
 787        called by the VM to write out pages associated with the
 788        address_space object.  If wbc->sync_mode is WB_SYNC_ALL, then
 789        the writeback_control will specify a range of pages that must be
 790        written out.  If it is WB_SYNC_NONE, then a nr_to_write is
 791        given and that many pages should be written if possible.  If no
 792        ->writepages is given, then mpage_writepages is used instead.
 793        This will choose pages from the address space that are tagged as
 794        DIRTY and will pass them to ->writepage.
 795
 796``set_page_dirty``
 797        called by the VM to set a page dirty.  This is particularly
 798        needed if an address space attaches private data to a page, and
 799        that data needs to be updated when a page is dirtied.  This is
 800        called, for example, when a memory mapped page gets modified.
 801        If defined, it should set the PageDirty flag, and the
 802        PAGECACHE_TAG_DIRTY tag in the radix tree.
 803
 804``readahead``
 805        Called by the VM to read pages associated with the address_space
 806        object.  The pages are consecutive in the page cache and are
 807        locked.  The implementation should decrement the page refcount
 808        after starting I/O on each page.  Usually the page will be
 809        unlocked by the I/O completion handler.  If the filesystem decides
 810        to stop attempting I/O before reaching the end of the readahead
 811        window, it can simply return.  The caller will decrement the page
 812        refcount and unlock the remaining pages for you.  Set PageUptodate
 813        if the I/O completes successfully.  Setting PageError on any page
 814        will be ignored; simply unlock the page if an I/O error occurs.
 815
 816``readpages``
 817        called by the VM to read pages associated with the address_space
 818        object.  This is essentially just a vector version of readpage.
 819        Instead of just one page, several pages are requested.
 820        readpages is only used for read-ahead, so read errors are
 821        ignored.  If anything goes wrong, feel free to give up.
 822        This interface is deprecated and will be removed by the end of
 823        2020; implement readahead instead.
 824
 825``write_begin``
 826        Called by the generic buffered write code to ask the filesystem
 827        to prepare to write len bytes at the given offset in the file.
 828        The address_space should check that the write will be able to
 829        complete, by allocating space if necessary and doing any other
 830        internal housekeeping.  If the write will update parts of any
 831        basic-blocks on storage, then those blocks should be pre-read
 832        (if they haven't been read already) so that the updated blocks
 833        can be written out properly.
 834
 835        The filesystem must return the locked pagecache page for the
 836        specified offset, in ``*pagep``, for the caller to write into.
 837
 838        It must be able to cope with short writes (where the length
 839        passed to write_begin is greater than the number of bytes copied
 840        into the page).
 841
 842        flags is a field for AOP_FLAG_xxx flags, described in
 843        include/linux/fs.h.
 844
 845        A void * may be returned in fsdata, which then gets passed into
 846        write_end.
 847
 848        Returns 0 on success; < 0 on failure (which is the error code),
 849        in which case write_end is not called.
 850
 851``write_end``
 852        After a successful write_begin, and data copy, write_end must be
 853        called.  len is the original len passed to write_begin, and
 854        copied is the amount that was able to be copied.
 855
 856        The filesystem must take care of unlocking the page and
 857        releasing it refcount, and updating i_size.
 858
 859        Returns < 0 on failure, otherwise the number of bytes (<=
 860        'copied') that were able to be copied into pagecache.
 861
 862``bmap``
 863        called by the VFS to map a logical block offset within object to
 864        physical block number.  This method is used by the FIBMAP ioctl
 865        and for working with swap-files.  To be able to swap to a file,
 866        the file must have a stable mapping to a block device.  The swap
 867        system does not go through the filesystem but instead uses bmap
 868        to find out where the blocks in the file are and uses those
 869        addresses directly.
 870
 871``invalidatepage``
 872        If a page has PagePrivate set, then invalidatepage will be
 873        called when part or all of the page is to be removed from the
 874        address space.  This generally corresponds to either a
 875        truncation, punch hole or a complete invalidation of the address
 876        space (in the latter case 'offset' will always be 0 and 'length'
 877        will be PAGE_SIZE).  Any private data associated with the page
 878        should be updated to reflect this truncation.  If offset is 0
 879        and length is PAGE_SIZE, then the private data should be
 880        released, because the page must be able to be completely
 881        discarded.  This may be done by calling the ->releasepage
 882        function, but in this case the release MUST succeed.
 883
 884``releasepage``
 885        releasepage is called on PagePrivate pages to indicate that the
 886        page should be freed if possible.  ->releasepage should remove
 887        any private data from the page and clear the PagePrivate flag.
 888        If releasepage() fails for some reason, it must indicate failure
 889        with a 0 return value.  releasepage() is used in two distinct
 890        though related cases.  The first is when the VM finds a clean
 891        page with no active users and wants to make it a free page.  If
 892        ->releasepage succeeds, the page will be removed from the
 893        address_space and become free.
 894
 895        The second case is when a request has been made to invalidate
 896        some or all pages in an address_space.  This can happen through
 897        the fadvise(POSIX_FADV_DONTNEED) system call or by the
 898        filesystem explicitly requesting it as nfs and 9fs do (when they
 899        believe the cache may be out of date with storage) by calling
 900        invalidate_inode_pages2().  If the filesystem makes such a call,
 901        and needs to be certain that all pages are invalidated, then its
 902        releasepage will need to ensure this.  Possibly it can clear the
 903        PageUptodate bit if it cannot free private data yet.
 904
 905``freepage``
 906        freepage is called once the page is no longer visible in the
 907        page cache in order to allow the cleanup of any private data.
 908        Since it may be called by the memory reclaimer, it should not
 909        assume that the original address_space mapping still exists, and
 910        it should not block.
 911
 912``direct_IO``
 913        called by the generic read/write routines to perform direct_IO -
 914        that is IO requests which bypass the page cache and transfer
 915        data directly between the storage and the application's address
 916        space.
 917
 918``isolate_page``
 919        Called by the VM when isolating a movable non-lru page.  If page
 920        is successfully isolated, VM marks the page as PG_isolated via
 921        __SetPageIsolated.
 922
 923``migrate_page``
 924        This is used to compact the physical memory usage.  If the VM
 925        wants to relocate a page (maybe off a memory card that is
 926        signalling imminent failure) it will pass a new page and an old
 927        page to this function.  migrate_page should transfer any private
 928        data across and update any references that it has to the page.
 929
 930``putback_page``
 931        Called by the VM when isolated page's migration fails.
 932
 933``launder_page``
 934        Called before freeing a page - it writes back the dirty page.
 935        To prevent redirtying the page, it is kept locked during the
 936        whole operation.
 937
 938``is_partially_uptodate``
 939        Called by the VM when reading a file through the pagecache when
 940        the underlying blocksize != pagesize.  If the required block is
 941        up to date then the read can complete without needing the IO to
 942        bring the whole page up to date.
 943
 944``is_dirty_writeback``
 945        Called by the VM when attempting to reclaim a page.  The VM uses
 946        dirty and writeback information to determine if it needs to
 947        stall to allow flushers a chance to complete some IO.
 948        Ordinarily it can use PageDirty and PageWriteback but some
 949        filesystems have more complex state (unstable pages in NFS
 950        prevent reclaim) or do not set those flags due to locking
 951        problems.  This callback allows a filesystem to indicate to the
 952        VM if a page should be treated as dirty or writeback for the
 953        purposes of stalling.
 954
 955``error_remove_page``
 956        normally set to generic_error_remove_page if truncation is ok
 957        for this address space.  Used for memory failure handling.
 958        Setting this implies you deal with pages going away under you,
 959        unless you have them locked or reference counts increased.
 960
 961``swap_activate``
 962        Called when swapon is used on a file to allocate space if
 963        necessary and pin the block lookup information in memory.  A
 964        return value of zero indicates success, in which case this file
 965        can be used to back swapspace.
 966
 967``swap_deactivate``
 968        Called during swapoff on files where swap_activate was
 969        successful.
 970
 971
 972The File Object
 973===============
 974
 975A file object represents a file opened by a process.  This is also known
 976as an "open file description" in POSIX parlance.
 977
 978
 979struct file_operations
 980----------------------
 981
 982This describes how the VFS can manipulate an open file.  As of kernel
 9834.18, the following members are defined:
 984
 985.. code-block:: c
 986
 987        struct file_operations {
 988                struct module *owner;
 989                loff_t (*llseek) (struct file *, loff_t, int);
 990                ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
 991                ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
 992                ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
 993                ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
 994                int (*iopoll)(struct kiocb *kiocb, bool spin);
 995                int (*iterate) (struct file *, struct dir_context *);
 996                int (*iterate_shared) (struct file *, struct dir_context *);
 997                __poll_t (*poll) (struct file *, struct poll_table_struct *);
 998                long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
 999                long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
1000                int (*mmap) (struct file *, struct vm_area_struct *);
1001                int (*open) (struct inode *, struct file *);
1002                int (*flush) (struct file *, fl_owner_t id);
1003                int (*release) (struct inode *, struct file *);
1004                int (*fsync) (struct file *, loff_t, loff_t, int datasync);
1005                int (*fasync) (int, struct file *, int);
1006                int (*lock) (struct file *, int, struct file_lock *);
1007                ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
1008                unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1009                int (*check_flags)(int);
1010                int (*flock) (struct file *, int, struct file_lock *);
1011                ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
1012                ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
1013                int (*setlease)(struct file *, long, struct file_lock **, void **);
1014                long (*fallocate)(struct file *file, int mode, loff_t offset,
1015                                  loff_t len);
1016                void (*show_fdinfo)(struct seq_file *m, struct file *f);
1017        #ifndef CONFIG_MMU
1018                unsigned (*mmap_capabilities)(struct file *);
1019        #endif
1020                ssize_t (*copy_file_range)(struct file *, loff_t, struct file *, loff_t, size_t, unsigned int);
1021                loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in,
1022                                           struct file *file_out, loff_t pos_out,
1023                                           loff_t len, unsigned int remap_flags);
1024                int (*fadvise)(struct file *, loff_t, loff_t, int);
1025        };
1026
1027Again, all methods are called without any locks being held, unless
1028otherwise noted.
1029
1030``llseek``
1031        called when the VFS needs to move the file position index
1032
1033``read``
1034        called by read(2) and related system calls
1035
1036``read_iter``
1037        possibly asynchronous read with iov_iter as destination
1038
1039``write``
1040        called by write(2) and related system calls
1041
1042``write_iter``
1043        possibly asynchronous write with iov_iter as source
1044
1045``iopoll``
1046        called when aio wants to poll for completions on HIPRI iocbs
1047
1048``iterate``
1049        called when the VFS needs to read the directory contents
1050
1051``iterate_shared``
1052        called when the VFS needs to read the directory contents when
1053        filesystem supports concurrent dir iterators
1054
1055``poll``
1056        called by the VFS when a process wants to check if there is
1057        activity on this file and (optionally) go to sleep until there
1058        is activity.  Called by the select(2) and poll(2) system calls
1059
1060``unlocked_ioctl``
1061        called by the ioctl(2) system call.
1062
1063``compat_ioctl``
1064        called by the ioctl(2) system call when 32 bit system calls are
1065         used on 64 bit kernels.
1066
1067``mmap``
1068        called by the mmap(2) system call
1069
1070``open``
1071        called by the VFS when an inode should be opened.  When the VFS
1072        opens a file, it creates a new "struct file".  It then calls the
1073        open method for the newly allocated file structure.  You might
1074        think that the open method really belongs in "struct
1075        inode_operations", and you may be right.  I think it's done the
1076        way it is because it makes filesystems simpler to implement.
1077        The open() method is a good place to initialize the
1078        "private_data" member in the file structure if you want to point
1079        to a device structure
1080
1081``flush``
1082        called by the close(2) system call to flush a file
1083
1084``release``
1085        called when the last reference to an open file is closed
1086
1087``fsync``
1088        called by the fsync(2) system call.  Also see the section above
1089        entitled "Handling errors during writeback".
1090
1091``fasync``
1092        called by the fcntl(2) system call when asynchronous
1093        (non-blocking) mode is enabled for a file
1094
1095``lock``
1096        called by the fcntl(2) system call for F_GETLK, F_SETLK, and
1097        F_SETLKW commands
1098
1099``get_unmapped_area``
1100        called by the mmap(2) system call
1101
1102``check_flags``
1103        called by the fcntl(2) system call for F_SETFL command
1104
1105``flock``
1106        called by the flock(2) system call
1107
1108``splice_write``
1109        called by the VFS to splice data from a pipe to a file.  This
1110        method is used by the splice(2) system call
1111
1112``splice_read``
1113        called by the VFS to splice data from file to a pipe.  This
1114        method is used by the splice(2) system call
1115
1116``setlease``
1117        called by the VFS to set or release a file lock lease.  setlease
1118        implementations should call generic_setlease to record or remove
1119        the lease in the inode after setting it.
1120
1121``fallocate``
1122        called by the VFS to preallocate blocks or punch a hole.
1123
1124``copy_file_range``
1125        called by the copy_file_range(2) system call.
1126
1127``remap_file_range``
1128        called by the ioctl(2) system call for FICLONERANGE and FICLONE
1129        and FIDEDUPERANGE commands to remap file ranges.  An
1130        implementation should remap len bytes at pos_in of the source
1131        file into the dest file at pos_out.  Implementations must handle
1132        callers passing in len == 0; this means "remap to the end of the
1133        source file".  The return value should the number of bytes
1134        remapped, or the usual negative error code if errors occurred
1135        before any bytes were remapped.  The remap_flags parameter
1136        accepts REMAP_FILE_* flags.  If REMAP_FILE_DEDUP is set then the
1137        implementation must only remap if the requested file ranges have
1138        identical contents.  If REMAP_FILE_CAN_SHORTEN is set, the caller is
1139        ok with the implementation shortening the request length to
1140        satisfy alignment or EOF requirements (or any other reason).
1141
1142``fadvise``
1143        possibly called by the fadvise64() system call.
1144
1145Note that the file operations are implemented by the specific
1146filesystem in which the inode resides.  When opening a device node
1147(character or block special) most filesystems will call special
1148support routines in the VFS which will locate the required device
1149driver information.  These support routines replace the filesystem file
1150operations with those for the device driver, and then proceed to call
1151the new open() method for the file.  This is how opening a device file
1152in the filesystem eventually ends up calling the device driver open()
1153method.
1154
1155
1156Directory Entry Cache (dcache)
1157==============================
1158
1159
1160struct dentry_operations
1161------------------------
1162
1163This describes how a filesystem can overload the standard dentry
1164operations.  Dentries and the dcache are the domain of the VFS and the
1165individual filesystem implementations.  Device drivers have no business
1166here.  These methods may be set to NULL, as they are either optional or
1167the VFS uses a default.  As of kernel 2.6.22, the following members are
1168defined:
1169
1170.. code-block:: c
1171
1172        struct dentry_operations {
1173                int (*d_revalidate)(struct dentry *, unsigned int);
1174                int (*d_weak_revalidate)(struct dentry *, unsigned int);
1175                int (*d_hash)(const struct dentry *, struct qstr *);
1176                int (*d_compare)(const struct dentry *,
1177                                 unsigned int, const char *, const struct qstr *);
1178                int (*d_delete)(const struct dentry *);
1179                int (*d_init)(struct dentry *);
1180                void (*d_release)(struct dentry *);
1181                void (*d_iput)(struct dentry *, struct inode *);
1182                char *(*d_dname)(struct dentry *, char *, int);
1183                struct vfsmount *(*d_automount)(struct path *);
1184                int (*d_manage)(const struct path *, bool);
1185                struct dentry *(*d_real)(struct dentry *, const struct inode *);
1186        };
1187
1188``d_revalidate``
1189        called when the VFS needs to revalidate a dentry.  This is
1190        called whenever a name look-up finds a dentry in the dcache.
1191        Most local filesystems leave this as NULL, because all their
1192        dentries in the dcache are valid.  Network filesystems are
1193        different since things can change on the server without the
1194        client necessarily being aware of it.
1195
1196        This function should return a positive value if the dentry is
1197        still valid, and zero or a negative error code if it isn't.
1198
1199        d_revalidate may be called in rcu-walk mode (flags &
1200        LOOKUP_RCU).  If in rcu-walk mode, the filesystem must
1201        revalidate the dentry without blocking or storing to the dentry,
1202        d_parent and d_inode should not be used without care (because
1203        they can change and, in d_inode case, even become NULL under
1204        us).
1205
1206        If a situation is encountered that rcu-walk cannot handle,
1207        return
1208        -ECHILD and it will be called again in ref-walk mode.
1209
1210``_weak_revalidate``
1211        called when the VFS needs to revalidate a "jumped" dentry.  This
1212        is called when a path-walk ends at dentry that was not acquired
1213        by doing a lookup in the parent directory.  This includes "/",
1214        "." and "..", as well as procfs-style symlinks and mountpoint
1215        traversal.
1216
1217        In this case, we are less concerned with whether the dentry is
1218        still fully correct, but rather that the inode is still valid.
1219        As with d_revalidate, most local filesystems will set this to
1220        NULL since their dcache entries are always valid.
1221
1222        This function has the same return code semantics as
1223        d_revalidate.
1224
1225        d_weak_revalidate is only called after leaving rcu-walk mode.
1226
1227``d_hash``
1228        called when the VFS adds a dentry to the hash table.  The first
1229        dentry passed to d_hash is the parent directory that the name is
1230        to be hashed into.
1231
1232        Same locking and synchronisation rules as d_compare regarding
1233        what is safe to dereference etc.
1234
1235``d_compare``
1236        called to compare a dentry name with a given name.  The first
1237        dentry is the parent of the dentry to be compared, the second is
1238        the child dentry.  len and name string are properties of the
1239        dentry to be compared.  qstr is the name to compare it with.
1240
1241        Must be constant and idempotent, and should not take locks if
1242        possible, and should not or store into the dentry.  Should not
1243        dereference pointers outside the dentry without lots of care
1244        (eg.  d_parent, d_inode, d_name should not be used).
1245
1246        However, our vfsmount is pinned, and RCU held, so the dentries
1247        and inodes won't disappear, neither will our sb or filesystem
1248        module.  ->d_sb may be used.
1249
1250        It is a tricky calling convention because it needs to be called
1251        under "rcu-walk", ie. without any locks or references on things.
1252
1253``d_delete``
1254        called when the last reference to a dentry is dropped and the
1255        dcache is deciding whether or not to cache it.  Return 1 to
1256        delete immediately, or 0 to cache the dentry.  Default is NULL
1257        which means to always cache a reachable dentry.  d_delete must
1258        be constant and idempotent.
1259
1260``d_init``
1261        called when a dentry is allocated
1262
1263``d_release``
1264        called when a dentry is really deallocated
1265
1266``d_iput``
1267        called when a dentry loses its inode (just prior to its being
1268        deallocated).  The default when this is NULL is that the VFS
1269        calls iput().  If you define this method, you must call iput()
1270        yourself
1271
1272``d_dname``
1273        called when the pathname of a dentry should be generated.
1274        Useful for some pseudo filesystems (sockfs, pipefs, ...) to
1275        delay pathname generation.  (Instead of doing it when dentry is
1276        created, it's done only when the path is needed.).  Real
1277        filesystems probably dont want to use it, because their dentries
1278        are present in global dcache hash, so their hash should be an
1279        invariant.  As no lock is held, d_dname() should not try to
1280        modify the dentry itself, unless appropriate SMP safety is used.
1281        CAUTION : d_path() logic is quite tricky.  The correct way to
1282        return for example "Hello" is to put it at the end of the
1283        buffer, and returns a pointer to the first char.
1284        dynamic_dname() helper function is provided to take care of
1285        this.
1286
1287        Example :
1288
1289.. code-block:: c
1290
1291        static char *pipefs_dname(struct dentry *dent, char *buffer, int buflen)
1292        {
1293                return dynamic_dname(dentry, buffer, buflen, "pipe:[%lu]",
1294                                dentry->d_inode->i_ino);
1295        }
1296
1297``d_automount``
1298        called when an automount dentry is to be traversed (optional).
1299        This should create a new VFS mount record and return the record
1300        to the caller.  The caller is supplied with a path parameter
1301        giving the automount directory to describe the automount target
1302        and the parent VFS mount record to provide inheritable mount
1303        parameters.  NULL should be returned if someone else managed to
1304        make the automount first.  If the vfsmount creation failed, then
1305        an error code should be returned.  If -EISDIR is returned, then
1306        the directory will be treated as an ordinary directory and
1307        returned to pathwalk to continue walking.
1308
1309        If a vfsmount is returned, the caller will attempt to mount it
1310        on the mountpoint and will remove the vfsmount from its
1311        expiration list in the case of failure.  The vfsmount should be
1312        returned with 2 refs on it to prevent automatic expiration - the
1313        caller will clean up the additional ref.
1314
1315        This function is only used if DCACHE_NEED_AUTOMOUNT is set on
1316        the dentry.  This is set by __d_instantiate() if S_AUTOMOUNT is
1317        set on the inode being added.
1318
1319``d_manage``
1320        called to allow the filesystem to manage the transition from a
1321        dentry (optional).  This allows autofs, for example, to hold up
1322        clients waiting to explore behind a 'mountpoint' while letting
1323        the daemon go past and construct the subtree there.  0 should be
1324        returned to let the calling process continue.  -EISDIR can be
1325        returned to tell pathwalk to use this directory as an ordinary
1326        directory and to ignore anything mounted on it and not to check
1327        the automount flag.  Any other error code will abort pathwalk
1328        completely.
1329
1330        If the 'rcu_walk' parameter is true, then the caller is doing a
1331        pathwalk in RCU-walk mode.  Sleeping is not permitted in this
1332        mode, and the caller can be asked to leave it and call again by
1333        returning -ECHILD.  -EISDIR may also be returned to tell
1334        pathwalk to ignore d_automount or any mounts.
1335
1336        This function is only used if DCACHE_MANAGE_TRANSIT is set on
1337        the dentry being transited from.
1338
1339``d_real``
1340        overlay/union type filesystems implement this method to return
1341        one of the underlying dentries hidden by the overlay.  It is
1342        used in two different modes:
1343
1344        Called from file_dentry() it returns the real dentry matching
1345        the inode argument.  The real dentry may be from a lower layer
1346        already copied up, but still referenced from the file.  This
1347        mode is selected with a non-NULL inode argument.
1348
1349        With NULL inode the topmost real underlying dentry is returned.
1350
1351Each dentry has a pointer to its parent dentry, as well as a hash list
1352of child dentries.  Child dentries are basically like files in a
1353directory.
1354
1355
1356Directory Entry Cache API
1357--------------------------
1358
1359There are a number of functions defined which permit a filesystem to
1360manipulate dentries:
1361
1362``dget``
1363        open a new handle for an existing dentry (this just increments
1364        the usage count)
1365
1366``dput``
1367        close a handle for a dentry (decrements the usage count).  If
1368        the usage count drops to 0, and the dentry is still in its
1369        parent's hash, the "d_delete" method is called to check whether
1370        it should be cached.  If it should not be cached, or if the
1371        dentry is not hashed, it is deleted.  Otherwise cached dentries
1372        are put into an LRU list to be reclaimed on memory shortage.
1373
1374``d_drop``
1375        this unhashes a dentry from its parents hash list.  A subsequent
1376        call to dput() will deallocate the dentry if its usage count
1377        drops to 0
1378
1379``d_delete``
1380        delete a dentry.  If there are no other open references to the
1381        dentry then the dentry is turned into a negative dentry (the
1382        d_iput() method is called).  If there are other references, then
1383        d_drop() is called instead
1384
1385``d_add``
1386        add a dentry to its parents hash list and then calls
1387        d_instantiate()
1388
1389``d_instantiate``
1390        add a dentry to the alias hash list for the inode and updates
1391        the "d_inode" member.  The "i_count" member in the inode
1392        structure should be set/incremented.  If the inode pointer is
1393        NULL, the dentry is called a "negative dentry".  This function
1394        is commonly called when an inode is created for an existing
1395        negative dentry
1396
1397``d_lookup``
1398        look up a dentry given its parent and path name component It
1399        looks up the child of that given name from the dcache hash
1400        table.  If it is found, the reference count is incremented and
1401        the dentry is returned.  The caller must use dput() to free the
1402        dentry when it finishes using it.
1403
1404
1405Mount Options
1406=============
1407
1408
1409Parsing options
1410---------------
1411
1412On mount and remount the filesystem is passed a string containing a
1413comma separated list of mount options.  The options can have either of
1414these forms:
1415
1416  option
1417  option=value
1418
1419The <linux/parser.h> header defines an API that helps parse these
1420options.  There are plenty of examples on how to use it in existing
1421filesystems.
1422
1423
1424Showing options
1425---------------
1426
1427If a filesystem accepts mount options, it must define show_options() to
1428show all the currently active options.  The rules are:
1429
1430  - options MUST be shown which are not default or their values differ
1431    from the default
1432
1433  - options MAY be shown which are enabled by default or have their
1434    default value
1435
1436Options used only internally between a mount helper and the kernel (such
1437as file descriptors), or which only have an effect during the mounting
1438(such as ones controlling the creation of a journal) are exempt from the
1439above rules.
1440
1441The underlying reason for the above rules is to make sure, that a mount
1442can be accurately replicated (e.g. umounting and mounting again) based
1443on the information found in /proc/mounts.
1444
1445
1446Resources
1447=========
1448
1449(Note some of these resources are not up-to-date with the latest kernel
1450 version.)
1451
1452Creating Linux virtual filesystems. 2002
1453    <https://lwn.net/Articles/13325/>
1454
1455The Linux Virtual File-system Layer by Neil Brown. 1999
1456    <http://www.cse.unsw.edu.au/~neilb/oss/linux-commentary/vfs.html>
1457
1458A tour of the Linux VFS by Michael K. Johnson. 1996
1459    <https://www.tldp.org/LDP/khg/HyperNews/get/fs/vfstour.html>
1460
1461A small trail through the Linux kernel by Andries Brouwer. 2001
1462    <https://www.win.tue.nl/~aeb/linux/vfs/trail.html>
1463