xref: /linux/Documentation/filesystems/vfs.rst (revision 52990390f91c1c39ca742fc8f390b29891d95127)
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.  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		int (*init_fs_context)(struct fs_context *);
119		const struct fs_parameter_spec *parameters;
120		struct dentry *(*mount) (struct file_system_type *, int,
121			const char *, void *);
122		void (*kill_sb) (struct super_block *);
123		struct module *owner;
124		struct file_system_type * next;
125		struct hlist_head fs_supers;
126
127		struct lock_class_key s_lock_key;
128		struct lock_class_key s_umount_key;
129		struct lock_class_key s_vfs_rename_key;
130		struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
131
132		struct lock_class_key i_lock_key;
133		struct lock_class_key i_mutex_key;
134		struct lock_class_key invalidate_lock_key;
135		struct lock_class_key i_mutex_dir_key;
136	};
137
138``name``
139	the name of the filesystem type, such as "ext2", "iso9660",
140	"msdos" and so on
141
142``fs_flags``
143	various flags (i.e. FS_REQUIRES_DEV, FS_NO_DCACHE, etc.)
144
145``init_fs_context``
146	Initializes 'struct fs_context' ->ops and ->fs_private fields with
147	filesystem-specific data.
148
149``parameters``
150	Pointer to the array of filesystem parameters descriptors
151	'struct fs_parameter_spec'.
152	More info in Documentation/filesystems/mount_api.rst.
153
154``mount``
155	the method to call when a new instance of this filesystem should
156	be mounted
157
158``kill_sb``
159	the method to call when an instance of this filesystem should be
160	shut down
161
162
163``owner``
164	for internal VFS use: you should initialize this to THIS_MODULE
165	in most cases.
166
167``next``
168	for internal VFS use: you should initialize this to NULL
169
170``fs_supers``
171	for internal VFS use: hlist of filesystem instances (superblocks)
172
173  s_lock_key, s_umount_key, s_vfs_rename_key, s_writers_key,
174  i_lock_key, i_mutex_key, invalidate_lock_key, i_mutex_dir_key: lockdep-specific
175
176The mount() method has the following arguments:
177
178``struct file_system_type *fs_type``
179	describes the filesystem, partly initialized by the specific
180	filesystem code
181
182``int flags``
183	mount flags
184
185``const char *dev_name``
186	the device name we are mounting.
187
188``void *data``
189	arbitrary mount options, usually comes as an ASCII string (see
190	"Mount Options" section)
191
192The mount() method must return the root dentry of the tree requested by
193caller.  An active reference to its superblock must be grabbed and the
194superblock must be locked.  On failure it should return ERR_PTR(error).
195
196The arguments match those of mount(2) and their interpretation depends
197on filesystem type.  E.g. for block filesystems, dev_name is interpreted
198as block device name, that device is opened and if it contains a
199suitable filesystem image the method creates and initializes struct
200super_block accordingly, returning its root dentry to caller.
201
202->mount() may choose to return a subtree of existing filesystem - it
203doesn't have to create a new one.  The main result from the caller's
204point of view is a reference to dentry at the root of (sub)tree to be
205attached; creation of new superblock is a common side effect.
206
207The most interesting member of the superblock structure that the mount()
208method fills in is the "s_op" field.  This is a pointer to a "struct
209super_operations" which describes the next level of the filesystem
210implementation.
211
212Usually, a filesystem uses one of the generic mount() implementations
213and provides a fill_super() callback instead.  The generic variants are:
214
215``mount_bdev``
216	mount a filesystem residing on a block device
217
218``mount_nodev``
219	mount a filesystem that is not backed by a device
220
221``mount_single``
222	mount a filesystem which shares the instance between all mounts
223
224A fill_super() callback implementation has the following arguments:
225
226``struct super_block *sb``
227	the superblock structure.  The callback must initialize this
228	properly.
229
230``void *data``
231	arbitrary mount options, usually comes as an ASCII string (see
232	"Mount Options" section)
233
234``int silent``
235	whether or not to be silent on error
236
237
238The Superblock Object
239=====================
240
241A superblock object represents a mounted filesystem.
242
243
244struct super_operations
245-----------------------
246
247This describes how the VFS can manipulate the superblock of your
248filesystem.  The following members are defined:
249
250.. code-block:: c
251
252	struct super_operations {
253		struct inode *(*alloc_inode)(struct super_block *sb);
254		void (*destroy_inode)(struct inode *);
255		void (*free_inode)(struct inode *);
256
257		void (*dirty_inode) (struct inode *, int flags);
258		int (*write_inode) (struct inode *, struct writeback_control *wbc);
259		int (*drop_inode) (struct inode *);
260		void (*evict_inode) (struct inode *);
261		void (*put_super) (struct super_block *);
262		int (*sync_fs)(struct super_block *sb, int wait);
263		int (*freeze_super) (struct super_block *);
264		int (*freeze_fs) (struct super_block *);
265		int (*thaw_super) (struct super_block *);
266		int (*unfreeze_fs) (struct super_block *);
267		int (*statfs) (struct dentry *, struct kstatfs *);
268		int (*remount_fs) (struct super_block *, int *, char *);
269		void (*umount_begin) (struct super_block *);
270
271		int (*show_options)(struct seq_file *, struct dentry *);
272		int (*show_devname)(struct seq_file *, struct dentry *);
273		int (*show_path)(struct seq_file *, struct dentry *);
274		int (*show_stats)(struct seq_file *, struct dentry *);
275
276		ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
277		ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
278		struct dquot **(*get_dquots)(struct inode *);
279
280		long (*nr_cached_objects)(struct super_block *,
281					struct shrink_control *);
282		long (*free_cached_objects)(struct super_block *,
283					struct shrink_control *);
284	};
285
286All methods are called without any locks being held, unless otherwise
287noted.  This means that most methods can block safely.  All methods are
288only called from a process context (i.e. not from an interrupt handler
289or bottom half).
290
291``alloc_inode``
292	this method is called by alloc_inode() to allocate memory for
293	struct inode and initialize it.  If this function is not
294	defined, a simple 'struct inode' is allocated.  Normally
295	alloc_inode will be used to allocate a larger structure which
296	contains a 'struct inode' embedded within it.
297
298``destroy_inode``
299	this method is called by destroy_inode() to release resources
300	allocated for struct inode.  It is only required if
301	->alloc_inode was defined and simply undoes anything done by
302	->alloc_inode.
303
304``free_inode``
305	this method is called from RCU callback. If you use call_rcu()
306	in ->destroy_inode to free 'struct inode' memory, then it's
307	better to release memory in this method.
308
309``dirty_inode``
310	this method is called by the VFS when an inode is marked dirty.
311	This is specifically for the inode itself being marked dirty,
312	not its data.  If the update needs to be persisted by fdatasync(),
313	then I_DIRTY_DATASYNC will be set in the flags argument.
314	I_DIRTY_TIME will be set in the flags in case lazytime is enabled
315	and struct inode has times updated since the last ->dirty_inode
316	call.
317
318``write_inode``
319	this method is called when the VFS needs to write an inode to
320	disc.  The second parameter indicates whether the write should
321	be synchronous or not, not all filesystems check this flag.
322
323``drop_inode``
324	called when the last access to the inode is dropped, with the
325	inode->i_lock spinlock held.
326
327	This method should be either NULL (normal UNIX filesystem
328	semantics) or "generic_delete_inode" (for filesystems that do
329	not want to cache inodes - causing "delete_inode" to always be
330	called regardless of the value of i_nlink)
331
332	The "generic_delete_inode()" behavior is equivalent to the old
333	practice of using "force_delete" in the put_inode() case, but
334	does not have the races that the "force_delete()" approach had.
335
336``evict_inode``
337	called when the VFS wants to evict an inode. Caller does
338	*not* evict the pagecache or inode-associated metadata buffers;
339	the method has to use truncate_inode_pages_final() to get rid
340	of those. Caller makes sure async writeback cannot be running for
341	the inode while (or after) ->evict_inode() is called. Optional.
342
343``put_super``
344	called when the VFS wishes to free the superblock
345	(i.e. unmount).  This is called with the superblock lock held
346
347``sync_fs``
348	called when VFS is writing out all dirty data associated with a
349	superblock.  The second parameter indicates whether the method
350	should wait until the write out has been completed.  Optional.
351
352``freeze_super``
353	Called instead of ->freeze_fs callback if provided.
354	Main difference is that ->freeze_super is called without taking
355	down_write(&sb->s_umount). If filesystem implements it and wants
356	->freeze_fs to be called too, then it has to call ->freeze_fs
357	explicitly from this callback. Optional.
358
359``freeze_fs``
360	called when VFS is locking a filesystem and forcing it into a
361	consistent state.  This method is currently used by the Logical
362	Volume Manager (LVM) and ioctl(FIFREEZE). Optional.
363
364``thaw_super``
365	called when VFS is unlocking a filesystem and making it writable
366	again after ->freeze_super. Optional.
367
368``unfreeze_fs``
369	called when VFS is unlocking a filesystem and making it writable
370	again after ->freeze_fs. Optional.
371
372``statfs``
373	called when the VFS needs to get filesystem statistics.
374
375``remount_fs``
376	called when the filesystem is remounted.  This is called with
377	the kernel lock held
378
379``umount_begin``
380	called when the VFS is unmounting a filesystem.
381
382``show_options``
383	called by the VFS to show mount options for /proc/<pid>/mounts
384	and /proc/<pid>/mountinfo.
385	(see "Mount Options" section)
386
387``show_devname``
388	Optional. Called by the VFS to show device name for
389	/proc/<pid>/{mounts,mountinfo,mountstats}. If not provided then
390	'(struct mount).mnt_devname' will be used.
391
392``show_path``
393	Optional. Called by the VFS (for /proc/<pid>/mountinfo) to show
394	the mount root dentry path relative to the filesystem root.
395
396``show_stats``
397	Optional. Called by the VFS (for /proc/<pid>/mountstats) to show
398	filesystem-specific mount statistics.
399
400``quota_read``
401	called by the VFS to read from filesystem quota file.
402
403``quota_write``
404	called by the VFS to write to filesystem quota file.
405
406``get_dquots``
407	called by quota to get 'struct dquot' array for a particular inode.
408	Optional.
409
410``nr_cached_objects``
411	called by the sb cache shrinking function for the filesystem to
412	return the number of freeable cached objects it contains.
413	Optional.
414
415``free_cache_objects``
416	called by the sb cache shrinking function for the filesystem to
417	scan the number of objects indicated to try to free them.
418	Optional, but any filesystem implementing this method needs to
419	also implement ->nr_cached_objects for it to be called
420	correctly.
421
422	We can't do anything with any errors that the filesystem might
423	encountered, hence the void return type.  This will never be
424	called if the VM is trying to reclaim under GFP_NOFS conditions,
425	hence this method does not need to handle that situation itself.
426
427	Implementations must include conditional reschedule calls inside
428	any scanning loop that is done.  This allows the VFS to
429	determine appropriate scan batch sizes without having to worry
430	about whether implementations will cause holdoff problems due to
431	large scan batch sizes.
432
433Whoever sets up the inode is responsible for filling in the "i_op"
434field.  This is a pointer to a "struct inode_operations" which describes
435the methods that can be performed on individual inodes.
436
437
438struct xattr_handlers
439---------------------
440
441On filesystems that support extended attributes (xattrs), the s_xattr
442superblock field points to a NULL-terminated array of xattr handlers.
443Extended attributes are name:value pairs.
444
445``name``
446	Indicates that the handler matches attributes with the specified
447	name (such as "system.posix_acl_access"); the prefix field must
448	be NULL.
449
450``prefix``
451	Indicates that the handler matches all attributes with the
452	specified name prefix (such as "user."); the name field must be
453	NULL.
454
455``list``
456	Determine if attributes matching this xattr handler should be
457	listed for a particular dentry.  Used by some listxattr
458	implementations like generic_listxattr.
459
460``get``
461	Called by the VFS to get the value of a particular extended
462	attribute.  This method is called by the getxattr(2) system
463	call.
464
465``set``
466	Called by the VFS to set the value of a particular extended
467	attribute.  When the new value is NULL, called to remove a
468	particular extended attribute.  This method is called by the
469	setxattr(2) and removexattr(2) system calls.
470
471When none of the xattr handlers of a filesystem match the specified
472attribute name or when a filesystem doesn't support extended attributes,
473the various ``*xattr(2)`` system calls return -EOPNOTSUPP.
474
475
476The Inode Object
477================
478
479An inode object represents an object within the filesystem.
480
481
482struct inode_operations
483-----------------------
484
485This describes how the VFS can manipulate an inode in your filesystem.
486As of kernel 2.6.22, the following members are defined:
487
488.. code-block:: c
489
490	struct inode_operations {
491		int (*create) (struct mnt_idmap *, struct inode *,struct dentry *, umode_t, bool);
492		struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
493		int (*link) (struct dentry *,struct inode *,struct dentry *);
494		int (*unlink) (struct inode *,struct dentry *);
495		int (*symlink) (struct mnt_idmap *, struct inode *,struct dentry *,const char *);
496		int (*mkdir) (struct mnt_idmap *, struct inode *,struct dentry *,umode_t);
497		int (*rmdir) (struct inode *,struct dentry *);
498		int (*mknod) (struct mnt_idmap *, struct inode *,struct dentry *,umode_t,dev_t);
499		int (*rename) (struct mnt_idmap *, struct inode *, struct dentry *,
500			       struct inode *, struct dentry *, unsigned int);
501		int (*readlink) (struct dentry *, char __user *,int);
502		const char *(*get_link) (struct dentry *, struct inode *,
503					 struct delayed_call *);
504		int (*permission) (struct mnt_idmap *, struct inode *, int);
505		struct posix_acl * (*get_inode_acl)(struct inode *, int, bool);
506		int (*setattr) (struct mnt_idmap *, struct dentry *, struct iattr *);
507		int (*getattr) (struct mnt_idmap *, const struct path *, struct kstat *, u32, unsigned int);
508		ssize_t (*listxattr) (struct dentry *, char *, size_t);
509		void (*update_time)(struct inode *, struct timespec *, int);
510		int (*atomic_open)(struct inode *, struct dentry *, struct file *,
511				   unsigned open_flag, umode_t create_mode);
512		int (*tmpfile) (struct mnt_idmap *, struct inode *, struct file *, umode_t);
513		struct posix_acl * (*get_acl)(struct mnt_idmap *, struct dentry *, int);
514	        int (*set_acl)(struct mnt_idmap *, struct dentry *, struct posix_acl *, int);
515		int (*fileattr_set)(struct mnt_idmap *idmap,
516				    struct dentry *dentry, struct fileattr *fa);
517		int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa);
518	};
519
520Again, all methods are called without any locks being held, unless
521otherwise noted.
522
523``create``
524	called by the open(2) and creat(2) system calls.  Only required
525	if you want to support regular files.  The dentry you get should
526	not have an inode (i.e. it should be a negative dentry).  Here
527	you will probably call d_instantiate() with the dentry and the
528	newly created inode
529
530``lookup``
531	called when the VFS needs to look up an inode in a parent
532	directory.  The name to look for is found in the dentry.  This
533	method must call d_add() to insert the found inode into the
534	dentry.  The "i_count" field in the inode structure should be
535	incremented.  If the named inode does not exist a NULL inode
536	should be inserted into the dentry (this is called a negative
537	dentry).  Returning an error code from this routine must only be
538	done on a real error, otherwise creating inodes with system
539	calls like create(2), mknod(2), mkdir(2) and so on will fail.
540	If you wish to overload the dentry methods then you should
541	initialise the "d_dop" field in the dentry; this is a pointer to
542	a struct "dentry_operations".  This method is called with the
543	directory inode semaphore held
544
545``link``
546	called by the link(2) system call.  Only required if you want to
547	support hard links.  You will probably need to call
548	d_instantiate() just as you would in the create() method
549
550``unlink``
551	called by the unlink(2) system call.  Only required if you want
552	to support deleting inodes
553
554``symlink``
555	called by the symlink(2) system call.  Only required if you want
556	to support symlinks.  You will probably need to call
557	d_instantiate() just as you would in the create() method
558
559``mkdir``
560	called by the mkdir(2) system call.  Only required if you want
561	to support creating subdirectories.  You will probably need to
562	call d_instantiate() just as you would in the create() method
563
564``rmdir``
565	called by the rmdir(2) system call.  Only required if you want
566	to support deleting subdirectories
567
568``mknod``
569	called by the mknod(2) system call to create a device (char,
570	block) inode or a named pipe (FIFO) or socket.  Only required if
571	you want to support creating these types of inodes.  You will
572	probably need to call d_instantiate() just as you would in the
573	create() method
574
575``rename``
576	called by the rename(2) system call to rename the object to have
577	the parent and name given by the second inode and dentry.
578
579	The filesystem must return -EINVAL for any unsupported or
580	unknown flags.  Currently the following flags are implemented:
581	(1) RENAME_NOREPLACE: this flag indicates that if the target of
582	the rename exists the rename should fail with -EEXIST instead of
583	replacing the target.  The VFS already checks for existence, so
584	for local filesystems the RENAME_NOREPLACE implementation is
585	equivalent to plain rename.
586	(2) RENAME_EXCHANGE: exchange source and target.  Both must
587	exist; this is checked by the VFS.  Unlike plain rename, source
588	and target may be of different type.
589
590``get_link``
591	called by the VFS to follow a symbolic link to the inode it
592	points to.  Only required if you want to support symbolic links.
593	This method returns the symlink body to traverse (and possibly
594	resets the current position with nd_jump_link()).  If the body
595	won't go away until the inode is gone, nothing else is needed;
596	if it needs to be otherwise pinned, arrange for its release by
597	having get_link(..., ..., done) do set_delayed_call(done,
598	destructor, argument).  In that case destructor(argument) will
599	be called once VFS is done with the body you've returned.  May
600	be called in RCU mode; that is indicated by NULL dentry
601	argument.  If request can't be handled without leaving RCU mode,
602	have it return ERR_PTR(-ECHILD).
603
604	If the filesystem stores the symlink target in ->i_link, the
605	VFS may use it directly without calling ->get_link(); however,
606	->get_link() must still be provided.  ->i_link must not be
607	freed until after an RCU grace period.  Writing to ->i_link
608	post-iget() time requires a 'release' memory barrier.
609
610``readlink``
611	this is now just an override for use by readlink(2) for the
612	cases when ->get_link uses nd_jump_link() or object is not in
613	fact a symlink.  Normally filesystems should only implement
614	->get_link for symlinks and readlink(2) will automatically use
615	that.
616
617``permission``
618	called by the VFS to check for access rights on a POSIX-like
619	filesystem.
620
621	May be called in rcu-walk mode (mask & MAY_NOT_BLOCK).  If in
622	rcu-walk mode, the filesystem must check the permission without
623	blocking or storing to the inode.
624
625	If a situation is encountered that rcu-walk cannot handle,
626	return
627	-ECHILD and it will be called again in ref-walk mode.
628
629``setattr``
630	called by the VFS to set attributes for a file.  This method is
631	called by chmod(2) and related system calls.
632
633``getattr``
634	called by the VFS to get attributes of a file.  This method is
635	called by stat(2) and related system calls.
636
637``listxattr``
638	called by the VFS to list all extended attributes for a given
639	file.  This method is called by the listxattr(2) system call.
640
641``update_time``
642	called by the VFS to update a specific time or the i_version of
643	an inode.  If this is not defined the VFS will update the inode
644	itself and call mark_inode_dirty_sync.
645
646``atomic_open``
647	called on the last component of an open.  Using this optional
648	method the filesystem can look up, possibly create and open the
649	file in one atomic operation.  If it wants to leave actual
650	opening to the caller (e.g. if the file turned out to be a
651	symlink, device, or just something filesystem won't do atomic
652	open for), it may signal this by returning finish_no_open(file,
653	dentry).  This method is only called if the last component is
654	negative or needs lookup.  Cached positive dentries are still
655	handled by f_op->open().  If the file was created, FMODE_CREATED
656	flag should be set in file->f_mode.  In case of O_EXCL the
657	method must only succeed if the file didn't exist and hence
658	FMODE_CREATED shall always be set on success.
659
660``tmpfile``
661	called in the end of O_TMPFILE open().  Optional, equivalent to
662	atomically creating, opening and unlinking a file in given
663	directory.  On success needs to return with the file already
664	open; this can be done by calling finish_open_simple() right at
665	the end.
666
667``fileattr_get``
668	called on ioctl(FS_IOC_GETFLAGS) and ioctl(FS_IOC_FSGETXATTR) to
669	retrieve miscellaneous file flags and attributes.  Also called
670	before the relevant SET operation to check what is being changed
671	(in this case with i_rwsem locked exclusive).  If unset, then
672	fall back to f_op->ioctl().
673
674``fileattr_set``
675	called on ioctl(FS_IOC_SETFLAGS) and ioctl(FS_IOC_FSSETXATTR) to
676	change miscellaneous file flags and attributes.  Callers hold
677	i_rwsem exclusive.  If unset, then fall back to f_op->ioctl().
678
679
680The Address Space Object
681========================
682
683The address space object is used to group and manage pages in the page
684cache.  It can be used to keep track of the pages in a file (or anything
685else) and also track the mapping of sections of the file into process
686address spaces.
687
688There are a number of distinct yet related services that an
689address-space can provide.  These include communicating memory pressure,
690page lookup by address, and keeping track of pages tagged as Dirty or
691Writeback.
692
693The first can be used independently to the others.  The VM can try to
694either write dirty pages in order to clean them, or release clean pages
695in order to reuse them.  To do this it can call the ->writepage method
696on dirty pages, and ->release_folio on clean folios with the private
697flag set.  Clean pages without PagePrivate and with no external references
698will be released without notice being given to the address_space.
699
700To achieve this functionality, pages need to be placed on an LRU with
701lru_cache_add and mark_page_active needs to be called whenever the page
702is used.
703
704Pages are normally kept in a radix tree index by ->index.  This tree
705maintains information about the PG_Dirty and PG_Writeback status of each
706page, so that pages with either of these flags can be found quickly.
707
708The Dirty tag is primarily used by mpage_writepages - the default
709->writepages method.  It uses the tag to find dirty pages to call
710->writepage on.  If mpage_writepages is not used (i.e. the address
711provides its own ->writepages) , the PAGECACHE_TAG_DIRTY tag is almost
712unused.  write_inode_now and sync_inode do use it (through
713__sync_single_inode) to check if ->writepages has been successful in
714writing out the whole address_space.
715
716The Writeback tag is used by filemap*wait* and sync_page* functions, via
717filemap_fdatawait_range, to wait for all writeback to complete.
718
719An address_space handler may attach extra information to a page,
720typically using the 'private' field in the 'struct page'.  If such
721information is attached, the PG_Private flag should be set.  This will
722cause various VM routines to make extra calls into the address_space
723handler to deal with that data.
724
725An address space acts as an intermediate between storage and
726application.  Data is read into the address space a whole page at a
727time, and provided to the application either by copying of the page, or
728by memory-mapping the page.  Data is written into the address space by
729the application, and then written-back to storage typically in whole
730pages, however the address_space has finer control of write sizes.
731
732The read process essentially only requires 'read_folio'.  The write
733process is more complicated and uses write_begin/write_end or
734dirty_folio to write data into the address_space, and writepage and
735writepages to writeback data to storage.
736
737Adding and removing pages to/from an address_space is protected by the
738inode's i_mutex.
739
740When data is written to a page, the PG_Dirty flag should be set.  It
741typically remains set until writepage asks for it to be written.  This
742should clear PG_Dirty and set PG_Writeback.  It can be actually written
743at any point after PG_Dirty is clear.  Once it is known to be safe,
744PG_Writeback is cleared.
745
746Writeback makes use of a writeback_control structure to direct the
747operations.  This gives the writepage and writepages operations some
748information about the nature of and reason for the writeback request,
749and the constraints under which it is being done.  It is also used to
750return information back to the caller about the result of a writepage or
751writepages request.
752
753
754Handling errors during writeback
755--------------------------------
756
757Most applications that do buffered I/O will periodically call a file
758synchronization call (fsync, fdatasync, msync or sync_file_range) to
759ensure that data written has made it to the backing store.  When there
760is an error during writeback, they expect that error to be reported when
761a file sync request is made.  After an error has been reported on one
762request, subsequent requests on the same file descriptor should return
7630, unless further writeback errors have occurred since the previous file
764syncronization.
765
766Ideally, the kernel would report errors only on file descriptions on
767which writes were done that subsequently failed to be written back.  The
768generic pagecache infrastructure does not track the file descriptions
769that have dirtied each individual page however, so determining which
770file descriptors should get back an error is not possible.
771
772Instead, the generic writeback error tracking infrastructure in the
773kernel settles for reporting errors to fsync on all file descriptions
774that were open at the time that the error occurred.  In a situation with
775multiple writers, all of them will get back an error on a subsequent
776fsync, even if all of the writes done through that particular file
777descriptor succeeded (or even if there were no writes on that file
778descriptor at all).
779
780Filesystems that wish to use this infrastructure should call
781mapping_set_error to record the error in the address_space when it
782occurs.  Then, after writing back data from the pagecache in their
783file->fsync operation, they should call file_check_and_advance_wb_err to
784ensure that the struct file's error cursor has advanced to the correct
785point in the stream of errors emitted by the backing device(s).
786
787
788struct address_space_operations
789-------------------------------
790
791This describes how the VFS can manipulate mapping of a file to page
792cache in your filesystem.  The following members are defined:
793
794.. code-block:: c
795
796	struct address_space_operations {
797		int (*writepage)(struct page *page, struct writeback_control *wbc);
798		int (*read_folio)(struct file *, struct folio *);
799		int (*writepages)(struct address_space *, struct writeback_control *);
800		bool (*dirty_folio)(struct address_space *, struct folio *);
801		void (*readahead)(struct readahead_control *);
802		int (*write_begin)(struct file *, struct address_space *mapping,
803				   loff_t pos, unsigned len,
804				struct page **pagep, void **fsdata);
805		int (*write_end)(struct file *, struct address_space *mapping,
806				 loff_t pos, unsigned len, unsigned copied,
807				 struct page *page, void *fsdata);
808		sector_t (*bmap)(struct address_space *, sector_t);
809		void (*invalidate_folio) (struct folio *, size_t start, size_t len);
810		bool (*release_folio)(struct folio *, gfp_t);
811		void (*free_folio)(struct folio *);
812		ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter);
813		int (*migrate_folio)(struct mapping *, struct folio *dst,
814				struct folio *src, enum migrate_mode);
815		int (*launder_folio) (struct folio *);
816
817		bool (*is_partially_uptodate) (struct folio *, size_t from,
818					       size_t count);
819		void (*is_dirty_writeback)(struct folio *, bool *, bool *);
820		int (*error_remove_page) (struct mapping *mapping, struct page *page);
821		int (*swap_activate)(struct swap_info_struct *sis, struct file *f, sector_t *span)
822		int (*swap_deactivate)(struct file *);
823		int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter);
824	};
825
826``writepage``
827	called by the VM to write a dirty page to backing store.  This
828	may happen for data integrity reasons (i.e. 'sync'), or to free
829	up memory (flush).  The difference can be seen in
830	wbc->sync_mode.  The PG_Dirty flag has been cleared and
831	PageLocked is true.  writepage should start writeout, should set
832	PG_Writeback, and should make sure the page is unlocked, either
833	synchronously or asynchronously when the write operation
834	completes.
835
836	If wbc->sync_mode is WB_SYNC_NONE, ->writepage doesn't have to
837	try too hard if there are problems, and may choose to write out
838	other pages from the mapping if that is easier (e.g. due to
839	internal dependencies).  If it chooses not to start writeout, it
840	should return AOP_WRITEPAGE_ACTIVATE so that the VM will not
841	keep calling ->writepage on that page.
842
843	See the file "Locking" for more details.
844
845``read_folio``
846	Called by the page cache to read a folio from the backing store.
847	The 'file' argument supplies authentication information to network
848	filesystems, and is generally not used by block based filesystems.
849	It may be NULL if the caller does not have an open file (eg if
850	the kernel is performing a read for itself rather than on behalf
851	of a userspace process with an open file).
852
853	If the mapping does not support large folios, the folio will
854	contain a single page.	The folio will be locked when read_folio
855	is called.  If the read completes successfully, the folio should
856	be marked uptodate.  The filesystem should unlock the folio
857	once the read has completed, whether it was successful or not.
858	The filesystem does not need to modify the refcount on the folio;
859	the page cache holds a reference count and that will not be
860	released until the folio is unlocked.
861
862	Filesystems may implement ->read_folio() synchronously.
863	In normal operation, folios are read through the ->readahead()
864	method.  Only if this fails, or if the caller needs to wait for
865	the read to complete will the page cache call ->read_folio().
866	Filesystems should not attempt to perform their own readahead
867	in the ->read_folio() operation.
868
869	If the filesystem cannot perform the read at this time, it can
870	unlock the folio, do whatever action it needs to ensure that the
871	read will succeed in the future and return AOP_TRUNCATED_PAGE.
872	In this case, the caller should look up the folio, lock it,
873	and call ->read_folio again.
874
875	Callers may invoke the ->read_folio() method directly, but using
876	read_mapping_folio() will take care of locking, waiting for the
877	read to complete and handle cases such as AOP_TRUNCATED_PAGE.
878
879``writepages``
880	called by the VM to write out pages associated with the
881	address_space object.  If wbc->sync_mode is WB_SYNC_ALL, then
882	the writeback_control will specify a range of pages that must be
883	written out.  If it is WB_SYNC_NONE, then a nr_to_write is
884	given and that many pages should be written if possible.  If no
885	->writepages is given, then mpage_writepages is used instead.
886	This will choose pages from the address space that are tagged as
887	DIRTY and will pass them to ->writepage.
888
889``dirty_folio``
890	called by the VM to mark a folio as dirty.  This is particularly
891	needed if an address space attaches private data to a folio, and
892	that data needs to be updated when a folio is dirtied.  This is
893	called, for example, when a memory mapped page gets modified.
894	If defined, it should set the folio dirty flag, and the
895	PAGECACHE_TAG_DIRTY search mark in i_pages.
896
897``readahead``
898	Called by the VM to read pages associated with the address_space
899	object.  The pages are consecutive in the page cache and are
900	locked.  The implementation should decrement the page refcount
901	after starting I/O on each page.  Usually the page will be
902	unlocked by the I/O completion handler.  The set of pages are
903	divided into some sync pages followed by some async pages,
904	rac->ra->async_size gives the number of async pages.  The
905	filesystem should attempt to read all sync pages but may decide
906	to stop once it reaches the async pages.  If it does decide to
907	stop attempting I/O, it can simply return.  The caller will
908	remove the remaining pages from the address space, unlock them
909	and decrement the page refcount.  Set PageUptodate if the I/O
910	completes successfully.  Setting PageError on any page will be
911	ignored; simply unlock the page if an I/O error occurs.
912
913``write_begin``
914	Called by the generic buffered write code to ask the filesystem
915	to prepare to write len bytes at the given offset in the file.
916	The address_space should check that the write will be able to
917	complete, by allocating space if necessary and doing any other
918	internal housekeeping.  If the write will update parts of any
919	basic-blocks on storage, then those blocks should be pre-read
920	(if they haven't been read already) so that the updated blocks
921	can be written out properly.
922
923	The filesystem must return the locked pagecache page for the
924	specified offset, in ``*pagep``, for the caller to write into.
925
926	It must be able to cope with short writes (where the length
927	passed to write_begin is greater than the number of bytes copied
928	into the page).
929
930	A void * may be returned in fsdata, which then gets passed into
931	write_end.
932
933	Returns 0 on success; < 0 on failure (which is the error code),
934	in which case write_end is not called.
935
936``write_end``
937	After a successful write_begin, and data copy, write_end must be
938	called.  len is the original len passed to write_begin, and
939	copied is the amount that was able to be copied.
940
941	The filesystem must take care of unlocking the page and
942	releasing it refcount, and updating i_size.
943
944	Returns < 0 on failure, otherwise the number of bytes (<=
945	'copied') that were able to be copied into pagecache.
946
947``bmap``
948	called by the VFS to map a logical block offset within object to
949	physical block number.  This method is used by the FIBMAP ioctl
950	and for working with swap-files.  To be able to swap to a file,
951	the file must have a stable mapping to a block device.  The swap
952	system does not go through the filesystem but instead uses bmap
953	to find out where the blocks in the file are and uses those
954	addresses directly.
955
956``invalidate_folio``
957	If a folio has private data, then invalidate_folio will be
958	called when part or all of the folio is to be removed from the
959	address space.  This generally corresponds to either a
960	truncation, punch hole or a complete invalidation of the address
961	space (in the latter case 'offset' will always be 0 and 'length'
962	will be folio_size()).  Any private data associated with the folio
963	should be updated to reflect this truncation.  If offset is 0
964	and length is folio_size(), then the private data should be
965	released, because the folio must be able to be completely
966	discarded.  This may be done by calling the ->release_folio
967	function, but in this case the release MUST succeed.
968
969``release_folio``
970	release_folio is called on folios with private data to tell the
971	filesystem that the folio is about to be freed.  ->release_folio
972	should remove any private data from the folio and clear the
973	private flag.  If release_folio() fails, it should return false.
974	release_folio() is used in two distinct though related cases.
975	The first is when the VM wants to free a clean folio with no
976	active users.  If ->release_folio succeeds, the folio will be
977	removed from the address_space and be freed.
978
979	The second case is when a request has been made to invalidate
980	some or all folios in an address_space.  This can happen
981	through the fadvise(POSIX_FADV_DONTNEED) system call or by the
982	filesystem explicitly requesting it as nfs and 9p do (when they
983	believe the cache may be out of date with storage) by calling
984	invalidate_inode_pages2().  If the filesystem makes such a call,
985	and needs to be certain that all folios are invalidated, then
986	its release_folio will need to ensure this.  Possibly it can
987	clear the uptodate flag if it cannot free private data yet.
988
989``free_folio``
990	free_folio is called once the folio is no longer visible in the
991	page cache in order to allow the cleanup of any private data.
992	Since it may be called by the memory reclaimer, it should not
993	assume that the original address_space mapping still exists, and
994	it should not block.
995
996``direct_IO``
997	called by the generic read/write routines to perform direct_IO -
998	that is IO requests which bypass the page cache and transfer
999	data directly between the storage and the application's address
1000	space.
1001
1002``migrate_folio``
1003	This is used to compact the physical memory usage.  If the VM
1004	wants to relocate a folio (maybe from a memory device that is
1005	signalling imminent failure) it will pass a new folio and an old
1006	folio to this function.  migrate_folio should transfer any private
1007	data across and update any references that it has to the folio.
1008
1009``launder_folio``
1010	Called before freeing a folio - it writes back the dirty folio.
1011	To prevent redirtying the folio, it is kept locked during the
1012	whole operation.
1013
1014``is_partially_uptodate``
1015	Called by the VM when reading a file through the pagecache when
1016	the underlying blocksize is smaller than the size of the folio.
1017	If the required block is up to date then the read can complete
1018	without needing I/O to bring the whole page up to date.
1019
1020``is_dirty_writeback``
1021	Called by the VM when attempting to reclaim a folio.  The VM uses
1022	dirty and writeback information to determine if it needs to
1023	stall to allow flushers a chance to complete some IO.
1024	Ordinarily it can use folio_test_dirty and folio_test_writeback but
1025	some filesystems have more complex state (unstable folios in NFS
1026	prevent reclaim) or do not set those flags due to locking
1027	problems.  This callback allows a filesystem to indicate to the
1028	VM if a folio should be treated as dirty or writeback for the
1029	purposes of stalling.
1030
1031``error_remove_page``
1032	normally set to generic_error_remove_page if truncation is ok
1033	for this address space.  Used for memory failure handling.
1034	Setting this implies you deal with pages going away under you,
1035	unless you have them locked or reference counts increased.
1036
1037``swap_activate``
1038
1039	Called to prepare the given file for swap.  It should perform
1040	any validation and preparation necessary to ensure that writes
1041	can be performed with minimal memory allocation.  It should call
1042	add_swap_extent(), or the helper iomap_swapfile_activate(), and
1043	return the number of extents added.  If IO should be submitted
1044	through ->swap_rw(), it should set SWP_FS_OPS, otherwise IO will
1045	be submitted directly to the block device ``sis->bdev``.
1046
1047``swap_deactivate``
1048	Called during swapoff on files where swap_activate was
1049	successful.
1050
1051``swap_rw``
1052	Called to read or write swap pages when SWP_FS_OPS is set.
1053
1054The File Object
1055===============
1056
1057A file object represents a file opened by a process.  This is also known
1058as an "open file description" in POSIX parlance.
1059
1060
1061struct file_operations
1062----------------------
1063
1064This describes how the VFS can manipulate an open file.  As of kernel
10654.18, the following members are defined:
1066
1067.. code-block:: c
1068
1069	struct file_operations {
1070		struct module *owner;
1071		loff_t (*llseek) (struct file *, loff_t, int);
1072		ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
1073		ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
1074		ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
1075		ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
1076		int (*iopoll)(struct kiocb *kiocb, bool spin);
1077		int (*iterate) (struct file *, struct dir_context *);
1078		int (*iterate_shared) (struct file *, struct dir_context *);
1079		__poll_t (*poll) (struct file *, struct poll_table_struct *);
1080		long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
1081		long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
1082		int (*mmap) (struct file *, struct vm_area_struct *);
1083		int (*open) (struct inode *, struct file *);
1084		int (*flush) (struct file *, fl_owner_t id);
1085		int (*release) (struct inode *, struct file *);
1086		int (*fsync) (struct file *, loff_t, loff_t, int datasync);
1087		int (*fasync) (int, struct file *, int);
1088		int (*lock) (struct file *, int, struct file_lock *);
1089		ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
1090		unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1091		int (*check_flags)(int);
1092		int (*flock) (struct file *, int, struct file_lock *);
1093		ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
1094		ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
1095		int (*setlease)(struct file *, long, struct file_lock **, void **);
1096		long (*fallocate)(struct file *file, int mode, loff_t offset,
1097				  loff_t len);
1098		void (*show_fdinfo)(struct seq_file *m, struct file *f);
1099	#ifndef CONFIG_MMU
1100		unsigned (*mmap_capabilities)(struct file *);
1101	#endif
1102		ssize_t (*copy_file_range)(struct file *, loff_t, struct file *, loff_t, size_t, unsigned int);
1103		loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in,
1104					   struct file *file_out, loff_t pos_out,
1105					   loff_t len, unsigned int remap_flags);
1106		int (*fadvise)(struct file *, loff_t, loff_t, int);
1107	};
1108
1109Again, all methods are called without any locks being held, unless
1110otherwise noted.
1111
1112``llseek``
1113	called when the VFS needs to move the file position index
1114
1115``read``
1116	called by read(2) and related system calls
1117
1118``read_iter``
1119	possibly asynchronous read with iov_iter as destination
1120
1121``write``
1122	called by write(2) and related system calls
1123
1124``write_iter``
1125	possibly asynchronous write with iov_iter as source
1126
1127``iopoll``
1128	called when aio wants to poll for completions on HIPRI iocbs
1129
1130``iterate``
1131	called when the VFS needs to read the directory contents
1132
1133``iterate_shared``
1134	called when the VFS needs to read the directory contents when
1135	filesystem supports concurrent dir iterators
1136
1137``poll``
1138	called by the VFS when a process wants to check if there is
1139	activity on this file and (optionally) go to sleep until there
1140	is activity.  Called by the select(2) and poll(2) system calls
1141
1142``unlocked_ioctl``
1143	called by the ioctl(2) system call.
1144
1145``compat_ioctl``
1146	called by the ioctl(2) system call when 32 bit system calls are
1147	 used on 64 bit kernels.
1148
1149``mmap``
1150	called by the mmap(2) system call
1151
1152``open``
1153	called by the VFS when an inode should be opened.  When the VFS
1154	opens a file, it creates a new "struct file".  It then calls the
1155	open method for the newly allocated file structure.  You might
1156	think that the open method really belongs in "struct
1157	inode_operations", and you may be right.  I think it's done the
1158	way it is because it makes filesystems simpler to implement.
1159	The open() method is a good place to initialize the
1160	"private_data" member in the file structure if you want to point
1161	to a device structure
1162
1163``flush``
1164	called by the close(2) system call to flush a file
1165
1166``release``
1167	called when the last reference to an open file is closed
1168
1169``fsync``
1170	called by the fsync(2) system call.  Also see the section above
1171	entitled "Handling errors during writeback".
1172
1173``fasync``
1174	called by the fcntl(2) system call when asynchronous
1175	(non-blocking) mode is enabled for a file
1176
1177``lock``
1178	called by the fcntl(2) system call for F_GETLK, F_SETLK, and
1179	F_SETLKW commands
1180
1181``get_unmapped_area``
1182	called by the mmap(2) system call
1183
1184``check_flags``
1185	called by the fcntl(2) system call for F_SETFL command
1186
1187``flock``
1188	called by the flock(2) system call
1189
1190``splice_write``
1191	called by the VFS to splice data from a pipe to a file.  This
1192	method is used by the splice(2) system call
1193
1194``splice_read``
1195	called by the VFS to splice data from file to a pipe.  This
1196	method is used by the splice(2) system call
1197
1198``setlease``
1199	called by the VFS to set or release a file lock lease.  setlease
1200	implementations should call generic_setlease to record or remove
1201	the lease in the inode after setting it.
1202
1203``fallocate``
1204	called by the VFS to preallocate blocks or punch a hole.
1205
1206``copy_file_range``
1207	called by the copy_file_range(2) system call.
1208
1209``remap_file_range``
1210	called by the ioctl(2) system call for FICLONERANGE and FICLONE
1211	and FIDEDUPERANGE commands to remap file ranges.  An
1212	implementation should remap len bytes at pos_in of the source
1213	file into the dest file at pos_out.  Implementations must handle
1214	callers passing in len == 0; this means "remap to the end of the
1215	source file".  The return value should the number of bytes
1216	remapped, or the usual negative error code if errors occurred
1217	before any bytes were remapped.  The remap_flags parameter
1218	accepts REMAP_FILE_* flags.  If REMAP_FILE_DEDUP is set then the
1219	implementation must only remap if the requested file ranges have
1220	identical contents.  If REMAP_FILE_CAN_SHORTEN is set, the caller is
1221	ok with the implementation shortening the request length to
1222	satisfy alignment or EOF requirements (or any other reason).
1223
1224``fadvise``
1225	possibly called by the fadvise64() system call.
1226
1227Note that the file operations are implemented by the specific
1228filesystem in which the inode resides.  When opening a device node
1229(character or block special) most filesystems will call special
1230support routines in the VFS which will locate the required device
1231driver information.  These support routines replace the filesystem file
1232operations with those for the device driver, and then proceed to call
1233the new open() method for the file.  This is how opening a device file
1234in the filesystem eventually ends up calling the device driver open()
1235method.
1236
1237
1238Directory Entry Cache (dcache)
1239==============================
1240
1241
1242struct dentry_operations
1243------------------------
1244
1245This describes how a filesystem can overload the standard dentry
1246operations.  Dentries and the dcache are the domain of the VFS and the
1247individual filesystem implementations.  Device drivers have no business
1248here.  These methods may be set to NULL, as they are either optional or
1249the VFS uses a default.  As of kernel 2.6.22, the following members are
1250defined:
1251
1252.. code-block:: c
1253
1254	struct dentry_operations {
1255		int (*d_revalidate)(struct dentry *, unsigned int);
1256		int (*d_weak_revalidate)(struct dentry *, unsigned int);
1257		int (*d_hash)(const struct dentry *, struct qstr *);
1258		int (*d_compare)(const struct dentry *,
1259				 unsigned int, const char *, const struct qstr *);
1260		int (*d_delete)(const struct dentry *);
1261		int (*d_init)(struct dentry *);
1262		void (*d_release)(struct dentry *);
1263		void (*d_iput)(struct dentry *, struct inode *);
1264		char *(*d_dname)(struct dentry *, char *, int);
1265		struct vfsmount *(*d_automount)(struct path *);
1266		int (*d_manage)(const struct path *, bool);
1267		struct dentry *(*d_real)(struct dentry *, const struct inode *);
1268	};
1269
1270``d_revalidate``
1271	called when the VFS needs to revalidate a dentry.  This is
1272	called whenever a name look-up finds a dentry in the dcache.
1273	Most local filesystems leave this as NULL, because all their
1274	dentries in the dcache are valid.  Network filesystems are
1275	different since things can change on the server without the
1276	client necessarily being aware of it.
1277
1278	This function should return a positive value if the dentry is
1279	still valid, and zero or a negative error code if it isn't.
1280
1281	d_revalidate may be called in rcu-walk mode (flags &
1282	LOOKUP_RCU).  If in rcu-walk mode, the filesystem must
1283	revalidate the dentry without blocking or storing to the dentry,
1284	d_parent and d_inode should not be used without care (because
1285	they can change and, in d_inode case, even become NULL under
1286	us).
1287
1288	If a situation is encountered that rcu-walk cannot handle,
1289	return
1290	-ECHILD and it will be called again in ref-walk mode.
1291
1292``d_weak_revalidate``
1293	called when the VFS needs to revalidate a "jumped" dentry.  This
1294	is called when a path-walk ends at dentry that was not acquired
1295	by doing a lookup in the parent directory.  This includes "/",
1296	"." and "..", as well as procfs-style symlinks and mountpoint
1297	traversal.
1298
1299	In this case, we are less concerned with whether the dentry is
1300	still fully correct, but rather that the inode is still valid.
1301	As with d_revalidate, most local filesystems will set this to
1302	NULL since their dcache entries are always valid.
1303
1304	This function has the same return code semantics as
1305	d_revalidate.
1306
1307	d_weak_revalidate is only called after leaving rcu-walk mode.
1308
1309``d_hash``
1310	called when the VFS adds a dentry to the hash table.  The first
1311	dentry passed to d_hash is the parent directory that the name is
1312	to be hashed into.
1313
1314	Same locking and synchronisation rules as d_compare regarding
1315	what is safe to dereference etc.
1316
1317``d_compare``
1318	called to compare a dentry name with a given name.  The first
1319	dentry is the parent of the dentry to be compared, the second is
1320	the child dentry.  len and name string are properties of the
1321	dentry to be compared.  qstr is the name to compare it with.
1322
1323	Must be constant and idempotent, and should not take locks if
1324	possible, and should not or store into the dentry.  Should not
1325	dereference pointers outside the dentry without lots of care
1326	(eg.  d_parent, d_inode, d_name should not be used).
1327
1328	However, our vfsmount is pinned, and RCU held, so the dentries
1329	and inodes won't disappear, neither will our sb or filesystem
1330	module.  ->d_sb may be used.
1331
1332	It is a tricky calling convention because it needs to be called
1333	under "rcu-walk", ie. without any locks or references on things.
1334
1335``d_delete``
1336	called when the last reference to a dentry is dropped and the
1337	dcache is deciding whether or not to cache it.  Return 1 to
1338	delete immediately, or 0 to cache the dentry.  Default is NULL
1339	which means to always cache a reachable dentry.  d_delete must
1340	be constant and idempotent.
1341
1342``d_init``
1343	called when a dentry is allocated
1344
1345``d_release``
1346	called when a dentry is really deallocated
1347
1348``d_iput``
1349	called when a dentry loses its inode (just prior to its being
1350	deallocated).  The default when this is NULL is that the VFS
1351	calls iput().  If you define this method, you must call iput()
1352	yourself
1353
1354``d_dname``
1355	called when the pathname of a dentry should be generated.
1356	Useful for some pseudo filesystems (sockfs, pipefs, ...) to
1357	delay pathname generation.  (Instead of doing it when dentry is
1358	created, it's done only when the path is needed.).  Real
1359	filesystems probably dont want to use it, because their dentries
1360	are present in global dcache hash, so their hash should be an
1361	invariant.  As no lock is held, d_dname() should not try to
1362	modify the dentry itself, unless appropriate SMP safety is used.
1363	CAUTION : d_path() logic is quite tricky.  The correct way to
1364	return for example "Hello" is to put it at the end of the
1365	buffer, and returns a pointer to the first char.
1366	dynamic_dname() helper function is provided to take care of
1367	this.
1368
1369	Example :
1370
1371.. code-block:: c
1372
1373	static char *pipefs_dname(struct dentry *dent, char *buffer, int buflen)
1374	{
1375		return dynamic_dname(dentry, buffer, buflen, "pipe:[%lu]",
1376				dentry->d_inode->i_ino);
1377	}
1378
1379``d_automount``
1380	called when an automount dentry is to be traversed (optional).
1381	This should create a new VFS mount record and return the record
1382	to the caller.  The caller is supplied with a path parameter
1383	giving the automount directory to describe the automount target
1384	and the parent VFS mount record to provide inheritable mount
1385	parameters.  NULL should be returned if someone else managed to
1386	make the automount first.  If the vfsmount creation failed, then
1387	an error code should be returned.  If -EISDIR is returned, then
1388	the directory will be treated as an ordinary directory and
1389	returned to pathwalk to continue walking.
1390
1391	If a vfsmount is returned, the caller will attempt to mount it
1392	on the mountpoint and will remove the vfsmount from its
1393	expiration list in the case of failure.  The vfsmount should be
1394	returned with 2 refs on it to prevent automatic expiration - the
1395	caller will clean up the additional ref.
1396
1397	This function is only used if DCACHE_NEED_AUTOMOUNT is set on
1398	the dentry.  This is set by __d_instantiate() if S_AUTOMOUNT is
1399	set on the inode being added.
1400
1401``d_manage``
1402	called to allow the filesystem to manage the transition from a
1403	dentry (optional).  This allows autofs, for example, to hold up
1404	clients waiting to explore behind a 'mountpoint' while letting
1405	the daemon go past and construct the subtree there.  0 should be
1406	returned to let the calling process continue.  -EISDIR can be
1407	returned to tell pathwalk to use this directory as an ordinary
1408	directory and to ignore anything mounted on it and not to check
1409	the automount flag.  Any other error code will abort pathwalk
1410	completely.
1411
1412	If the 'rcu_walk' parameter is true, then the caller is doing a
1413	pathwalk in RCU-walk mode.  Sleeping is not permitted in this
1414	mode, and the caller can be asked to leave it and call again by
1415	returning -ECHILD.  -EISDIR may also be returned to tell
1416	pathwalk to ignore d_automount or any mounts.
1417
1418	This function is only used if DCACHE_MANAGE_TRANSIT is set on
1419	the dentry being transited from.
1420
1421``d_real``
1422	overlay/union type filesystems implement this method to return
1423	one of the underlying dentries hidden by the overlay.  It is
1424	used in two different modes:
1425
1426	Called from file_dentry() it returns the real dentry matching
1427	the inode argument.  The real dentry may be from a lower layer
1428	already copied up, but still referenced from the file.  This
1429	mode is selected with a non-NULL inode argument.
1430
1431	With NULL inode the topmost real underlying dentry is returned.
1432
1433Each dentry has a pointer to its parent dentry, as well as a hash list
1434of child dentries.  Child dentries are basically like files in a
1435directory.
1436
1437
1438Directory Entry Cache API
1439--------------------------
1440
1441There are a number of functions defined which permit a filesystem to
1442manipulate dentries:
1443
1444``dget``
1445	open a new handle for an existing dentry (this just increments
1446	the usage count)
1447
1448``dput``
1449	close a handle for a dentry (decrements the usage count).  If
1450	the usage count drops to 0, and the dentry is still in its
1451	parent's hash, the "d_delete" method is called to check whether
1452	it should be cached.  If it should not be cached, or if the
1453	dentry is not hashed, it is deleted.  Otherwise cached dentries
1454	are put into an LRU list to be reclaimed on memory shortage.
1455
1456``d_drop``
1457	this unhashes a dentry from its parents hash list.  A subsequent
1458	call to dput() will deallocate the dentry if its usage count
1459	drops to 0
1460
1461``d_delete``
1462	delete a dentry.  If there are no other open references to the
1463	dentry then the dentry is turned into a negative dentry (the
1464	d_iput() method is called).  If there are other references, then
1465	d_drop() is called instead
1466
1467``d_add``
1468	add a dentry to its parents hash list and then calls
1469	d_instantiate()
1470
1471``d_instantiate``
1472	add a dentry to the alias hash list for the inode and updates
1473	the "d_inode" member.  The "i_count" member in the inode
1474	structure should be set/incremented.  If the inode pointer is
1475	NULL, the dentry is called a "negative dentry".  This function
1476	is commonly called when an inode is created for an existing
1477	negative dentry
1478
1479``d_lookup``
1480	look up a dentry given its parent and path name component It
1481	looks up the child of that given name from the dcache hash
1482	table.  If it is found, the reference count is incremented and
1483	the dentry is returned.  The caller must use dput() to free the
1484	dentry when it finishes using it.
1485
1486
1487Mount Options
1488=============
1489
1490
1491Parsing options
1492---------------
1493
1494On mount and remount the filesystem is passed a string containing a
1495comma separated list of mount options.  The options can have either of
1496these forms:
1497
1498  option
1499  option=value
1500
1501The <linux/parser.h> header defines an API that helps parse these
1502options.  There are plenty of examples on how to use it in existing
1503filesystems.
1504
1505
1506Showing options
1507---------------
1508
1509If a filesystem accepts mount options, it must define show_options() to
1510show all the currently active options.  The rules are:
1511
1512  - options MUST be shown which are not default or their values differ
1513    from the default
1514
1515  - options MAY be shown which are enabled by default or have their
1516    default value
1517
1518Options used only internally between a mount helper and the kernel (such
1519as file descriptors), or which only have an effect during the mounting
1520(such as ones controlling the creation of a journal) are exempt from the
1521above rules.
1522
1523The underlying reason for the above rules is to make sure, that a mount
1524can be accurately replicated (e.g. umounting and mounting again) based
1525on the information found in /proc/mounts.
1526
1527
1528Resources
1529=========
1530
1531(Note some of these resources are not up-to-date with the latest kernel
1532 version.)
1533
1534Creating Linux virtual filesystems. 2002
1535    <https://lwn.net/Articles/13325/>
1536
1537The Linux Virtual File-system Layer by Neil Brown. 1999
1538    <http://www.cse.unsw.edu.au/~neilb/oss/linux-commentary/vfs.html>
1539
1540A tour of the Linux VFS by Michael K. Johnson. 1996
1541    <https://www.tldp.org/LDP/khg/HyperNews/get/fs/vfstour.html>
1542
1543A small trail through the Linux kernel by Andries Brouwer. 2001
1544    <https://www.win.tue.nl/~aeb/linux/vfs/trail.html>
1545