1 .. SPDX-License-Identifier: GPL-2.0
3 =========================================
4 Overview of the Linux Virtual File System
5 =========================================
7 Original author: Richard Gooch <rgooch@atnf.csiro.au>
9 - Copyright (C) 1999 Richard Gooch
10 - Copyright (C) 2005 Pekka Enberg
16 The Virtual File System (also known as the Virtual Filesystem Switch) is
17 the software layer in the kernel that provides the filesystem interface
18 to userspace programs. It also provides an abstraction within the
19 kernel which allows different filesystem implementations to coexist.
21 VFS system calls open(2), stat(2), read(2), write(2), chmod(2) and so on
22 are called from a process context. Filesystem locking is described in
23 the document Documentation/filesystems/locking.rst.
26 Directory Entry Cache (dcache)
27 ------------------------------
29 The VFS implements the open(2), stat(2), chmod(2), and similar system
30 calls. The pathname argument that is passed to them is used by the VFS
31 to search through the directory entry cache (also known as the dentry
32 cache or dcache). This provides a very fast look-up mechanism to
33 translate a pathname (filename) into a specific dentry. Dentries live
34 in RAM and are never saved to disc: they exist only for performance.
36 The dentry cache is meant to be a view into your entire filespace. As
37 most computers cannot fit all dentries in the RAM at the same time, some
38 bits of the cache are missing. In order to resolve your pathname into a
39 dentry, the VFS may have to resort to creating dentries along the way,
40 and then loading the inode. This is done by looking up the inode.
46 An individual dentry usually has a pointer to an inode. Inodes are
47 filesystem objects such as regular files, directories, FIFOs and other
48 beasts. They live either on the disc (for block device filesystems) or
49 in the memory (for pseudo filesystems). Inodes that live on the disc
50 are copied into the memory when required and changes to the inode are
51 written back to disc. A single inode can be pointed to by multiple
52 dentries (hard links, for example, do this).
54 To look up an inode requires that the VFS calls the lookup() method of
55 the parent directory inode. This method is installed by the specific
56 filesystem implementation that the inode lives in. Once the VFS has the
57 required dentry (and hence the inode), we can do all those boring things
58 like open(2) the file, or stat(2) it to peek at the inode data. The
59 stat(2) operation is fairly simple: once the VFS has the dentry, it
60 peeks at the inode data and passes some of it back to userspace.
66 Opening a file requires another operation: allocation of a file
67 structure (this is the kernel-side implementation of file descriptors).
68 The freshly allocated file structure is initialized with a pointer to
69 the dentry and a set of file operation member functions. These are
70 taken from the inode data. The open() file method is then called so the
71 specific filesystem implementation can do its work. You can see that
72 this is another switch performed by the VFS. The file structure is
73 placed into the file descriptor table for the process.
75 Reading, writing and closing files (and other assorted VFS operations)
76 is done by using the userspace file descriptor to grab the appropriate
77 file structure, and then calling the required file structure method to
78 do whatever is required. For as long as the file is open, it keeps the
79 dentry in use, which in turn means that the VFS inode is still in use.
82 Registering and Mounting a Filesystem
83 =====================================
85 To register and unregister a filesystem, use the following API
92 extern int register_filesystem(struct file_system_type *);
93 extern int unregister_filesystem(struct file_system_type *);
95 The passed struct file_system_type describes your filesystem. When a
96 request is made to mount a filesystem onto a directory in your
97 namespace, the VFS will call the appropriate mount() method for the
98 specific filesystem. New vfsmount referring to the tree returned by
99 ->mount() will be attached to the mountpoint, so that when pathname
100 resolution reaches the mountpoint it will jump into the root of that
103 You can see all filesystems that are registered to the kernel in the
104 file /proc/filesystems.
107 struct file_system_type
108 -----------------------
110 This describes the filesystem. As of kernel 2.6.39, the following
115 struct file_system_operations {
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;
129 the name of the filesystem type, such as "ext2", "iso9660",
133 various flags (i.e. FS_REQUIRES_DEV, FS_NO_DCACHE, etc.)
136 the method to call when a new instance of this filesystem should
140 the method to call when an instance of this filesystem should be
145 for internal VFS use: you should initialize this to THIS_MODULE
149 for internal VFS use: you should initialize this to NULL
151 s_lock_key, s_umount_key: lockdep-specific
153 The mount() method has the following arguments:
155 ``struct file_system_type *fs_type``
156 describes the filesystem, partly initialized by the specific
162 ``const char *dev_name``
163 the device name we are mounting.
166 arbitrary mount options, usually comes as an ASCII string (see
167 "Mount Options" section)
169 The mount() method must return the root dentry of the tree requested by
170 caller. An active reference to its superblock must be grabbed and the
171 superblock must be locked. On failure it should return ERR_PTR(error).
173 The arguments match those of mount(2) and their interpretation depends
174 on filesystem type. E.g. for block filesystems, dev_name is interpreted
175 as block device name, that device is opened and if it contains a
176 suitable filesystem image the method creates and initializes struct
177 super_block accordingly, returning its root dentry to caller.
179 ->mount() may choose to return a subtree of existing filesystem - it
180 doesn't have to create a new one. The main result from the caller's
181 point of view is a reference to dentry at the root of (sub)tree to be
182 attached; creation of new superblock is a common side effect.
184 The most interesting member of the superblock structure that the mount()
185 method fills in is the "s_op" field. This is a pointer to a "struct
186 super_operations" which describes the next level of the filesystem
189 Usually, a filesystem uses one of the generic mount() implementations
190 and provides a fill_super() callback instead. The generic variants are:
193 mount a filesystem residing on a block device
196 mount a filesystem that is not backed by a device
199 mount a filesystem which shares the instance between all mounts
201 A fill_super() callback implementation has the following arguments:
203 ``struct super_block *sb``
204 the superblock structure. The callback must initialize this
208 arbitrary mount options, usually comes as an ASCII string (see
209 "Mount Options" section)
212 whether or not to be silent on error
215 The Superblock Object
216 =====================
218 A superblock object represents a mounted filesystem.
221 struct super_operations
222 -----------------------
224 This describes how the VFS can manipulate the superblock of your
225 filesystem. As of kernel 2.6.22, the following members are defined:
229 struct super_operations {
230 struct inode *(*alloc_inode)(struct super_block *sb);
231 void (*destroy_inode)(struct inode *);
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 *);
246 int (*show_options)(struct seq_file *, struct dentry *);
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);
254 All methods are called without any locks being held, unless otherwise
255 noted. This means that most methods can block safely. All methods are
256 only called from a process context (i.e. not from an interrupt handler
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.
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
273 this method is called by the VFS to mark an inode dirty.
276 this method is called when the VFS needs to write an inode to
277 disc. The second parameter indicates whether the write should
278 be synchronous or not, not all filesystems check this flag.
281 called when the last access to the inode is dropped, with the
282 inode->i_lock spinlock held.
284 This method should be either NULL (normal UNIX filesystem
285 semantics) or "generic_delete_inode" (for filesystems that do
286 not want to cache inodes - causing "delete_inode" to always be
287 called regardless of the value of i_nlink)
289 The "generic_delete_inode()" behavior is equivalent to the old
290 practice of using "force_delete" in the put_inode() case, but
291 does not have the races that the "force_delete()" approach had.
294 called when the VFS wants to delete an inode
297 called when the VFS wishes to free the superblock
298 (i.e. unmount). This is called with the superblock lock held
301 called when VFS is writing out all dirty data associated with a
302 superblock. The second parameter indicates whether the method
303 should wait until the write out has been completed. Optional.
306 called when VFS is locking a filesystem and forcing it into a
307 consistent state. This method is currently used by the Logical
308 Volume Manager (LVM).
311 called when VFS is unlocking a filesystem and making it writable
315 called when the VFS needs to get filesystem statistics.
318 called when the filesystem is remounted. This is called with
322 called then the VFS clears the inode. Optional
325 called when the VFS is unmounting a filesystem.
328 called by the VFS to show mount options for /proc/<pid>/mounts.
329 (see "Mount Options" section)
332 called by the VFS to read from filesystem quota file.
335 called by the VFS to write to filesystem quota file.
337 ``nr_cached_objects``
338 called by the sb cache shrinking function for the filesystem to
339 return the number of freeable cached objects it contains.
342 ``free_cache_objects``
343 called by the sb cache shrinking function for the filesystem to
344 scan the number of objects indicated to try to free them.
345 Optional, but any filesystem implementing this method needs to
346 also implement ->nr_cached_objects for it to be called
349 We can't do anything with any errors that the filesystem might
350 encountered, hence the void return type. This will never be
351 called if the VM is trying to reclaim under GFP_NOFS conditions,
352 hence this method does not need to handle that situation itself.
354 Implementations must include conditional reschedule calls inside
355 any scanning loop that is done. This allows the VFS to
356 determine appropriate scan batch sizes without having to worry
357 about whether implementations will cause holdoff problems due to
358 large scan batch sizes.
360 Whoever sets up the inode is responsible for filling in the "i_op"
361 field. This is a pointer to a "struct inode_operations" which describes
362 the methods that can be performed on individual inodes.
365 struct xattr_handlers
366 ---------------------
368 On filesystems that support extended attributes (xattrs), the s_xattr
369 superblock field points to a NULL-terminated array of xattr handlers.
370 Extended attributes are name:value pairs.
373 Indicates that the handler matches attributes with the specified
374 name (such as "system.posix_acl_access"); the prefix field must
378 Indicates that the handler matches all attributes with the
379 specified name prefix (such as "user."); the name field must be
383 Determine if attributes matching this xattr handler should be
384 listed for a particular dentry. Used by some listxattr
385 implementations like generic_listxattr.
388 Called by the VFS to get the value of a particular extended
389 attribute. This method is called by the getxattr(2) system
393 Called by the VFS to set the value of a particular extended
394 attribute. When the new value is NULL, called to remove a
395 particular extended attribute. This method is called by the the
396 setxattr(2) and removexattr(2) system calls.
398 When none of the xattr handlers of a filesystem match the specified
399 attribute name or when a filesystem doesn't support extended attributes,
400 the various ``*xattr(2)`` system calls return -EOPNOTSUPP.
406 An inode object represents an object within the filesystem.
409 struct inode_operations
410 -----------------------
412 This describes how the VFS can manipulate an inode in your filesystem.
413 As of kernel 2.6.22, the following members are defined:
417 struct inode_operations {
418 int (*create) (struct inode *,struct dentry *, umode_t, bool);
419 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
420 int (*link) (struct dentry *,struct inode *,struct dentry *);
421 int (*unlink) (struct inode *,struct dentry *);
422 int (*symlink) (struct inode *,struct dentry *,const char *);
423 int (*mkdir) (struct inode *,struct dentry *,umode_t);
424 int (*rmdir) (struct inode *,struct dentry *);
425 int (*mknod) (struct inode *,struct dentry *,umode_t,dev_t);
426 int (*rename) (struct inode *, struct dentry *,
427 struct inode *, struct dentry *, unsigned int);
428 int (*readlink) (struct dentry *, char __user *,int);
429 const char *(*get_link) (struct dentry *, struct inode *,
430 struct delayed_call *);
431 int (*permission) (struct inode *, int);
432 int (*get_acl)(struct inode *, int);
433 int (*setattr) (struct dentry *, struct iattr *);
434 int (*getattr) (const struct path *, struct kstat *, u32, unsigned int);
435 ssize_t (*listxattr) (struct dentry *, char *, size_t);
436 void (*update_time)(struct inode *, struct timespec *, int);
437 int (*atomic_open)(struct inode *, struct dentry *, struct file *,
438 unsigned open_flag, umode_t create_mode);
439 int (*tmpfile) (struct inode *, struct dentry *, umode_t);
442 Again, all methods are called without any locks being held, unless
446 called by the open(2) and creat(2) system calls. Only required
447 if you want to support regular files. The dentry you get should
448 not have an inode (i.e. it should be a negative dentry). Here
449 you will probably call d_instantiate() with the dentry and the
453 called when the VFS needs to look up an inode in a parent
454 directory. The name to look for is found in the dentry. This
455 method must call d_add() to insert the found inode into the
456 dentry. The "i_count" field in the inode structure should be
457 incremented. If the named inode does not exist a NULL inode
458 should be inserted into the dentry (this is called a negative
459 dentry). Returning an error code from this routine must only be
460 done on a real error, otherwise creating inodes with system
461 calls like create(2), mknod(2), mkdir(2) and so on will fail.
462 If you wish to overload the dentry methods then you should
463 initialise the "d_dop" field in the dentry; this is a pointer to
464 a struct "dentry_operations". This method is called with the
465 directory inode semaphore held
468 called by the link(2) system call. Only required if you want to
469 support hard links. You will probably need to call
470 d_instantiate() just as you would in the create() method
473 called by the unlink(2) system call. Only required if you want
474 to support deleting inodes
477 called by the symlink(2) system call. Only required if you want
478 to support symlinks. You will probably need to call
479 d_instantiate() just as you would in the create() method
482 called by the mkdir(2) system call. Only required if you want
483 to support creating subdirectories. You will probably need to
484 call d_instantiate() just as you would in the create() method
487 called by the rmdir(2) system call. Only required if you want
488 to support deleting subdirectories
491 called by the mknod(2) system call to create a device (char,
492 block) inode or a named pipe (FIFO) or socket. Only required if
493 you want to support creating these types of inodes. You will
494 probably need to call d_instantiate() just as you would in the
498 called by the rename(2) system call to rename the object to have
499 the parent and name given by the second inode and dentry.
501 The filesystem must return -EINVAL for any unsupported or
502 unknown flags. Currently the following flags are implemented:
503 (1) RENAME_NOREPLACE: this flag indicates that if the target of
504 the rename exists the rename should fail with -EEXIST instead of
505 replacing the target. The VFS already checks for existence, so
506 for local filesystems the RENAME_NOREPLACE implementation is
507 equivalent to plain rename.
508 (2) RENAME_EXCHANGE: exchange source and target. Both must
509 exist; this is checked by the VFS. Unlike plain rename, source
510 and target may be of different type.
513 called by the VFS to follow a symbolic link to the inode it
514 points to. Only required if you want to support symbolic links.
515 This method returns the symlink body to traverse (and possibly
516 resets the current position with nd_jump_link()). If the body
517 won't go away until the inode is gone, nothing else is needed;
518 if it needs to be otherwise pinned, arrange for its release by
519 having get_link(..., ..., done) do set_delayed_call(done,
520 destructor, argument). In that case destructor(argument) will
521 be called once VFS is done with the body you've returned. May
522 be called in RCU mode; that is indicated by NULL dentry
523 argument. If request can't be handled without leaving RCU mode,
524 have it return ERR_PTR(-ECHILD).
526 If the filesystem stores the symlink target in ->i_link, the
527 VFS may use it directly without calling ->get_link(); however,
528 ->get_link() must still be provided. ->i_link must not be
529 freed until after an RCU grace period. Writing to ->i_link
530 post-iget() time requires a 'release' memory barrier.
533 this is now just an override for use by readlink(2) for the
534 cases when ->get_link uses nd_jump_link() or object is not in
535 fact a symlink. Normally filesystems should only implement
536 ->get_link for symlinks and readlink(2) will automatically use
540 called by the VFS to check for access rights on a POSIX-like
543 May be called in rcu-walk mode (mask & MAY_NOT_BLOCK). If in
544 rcu-walk mode, the filesystem must check the permission without
545 blocking or storing to the inode.
547 If a situation is encountered that rcu-walk cannot handle,
549 -ECHILD and it will be called again in ref-walk mode.
552 called by the VFS to set attributes for a file. This method is
553 called by chmod(2) and related system calls.
556 called by the VFS to get attributes of a file. This method is
557 called by stat(2) and related system calls.
560 called by the VFS to list all extended attributes for a given
561 file. This method is called by the listxattr(2) system call.
564 called by the VFS to update a specific time or the i_version of
565 an inode. If this is not defined the VFS will update the inode
566 itself and call mark_inode_dirty_sync.
569 called on the last component of an open. Using this optional
570 method the filesystem can look up, possibly create and open the
571 file in one atomic operation. If it wants to leave actual
572 opening to the caller (e.g. if the file turned out to be a
573 symlink, device, or just something filesystem won't do atomic
574 open for), it may signal this by returning finish_no_open(file,
575 dentry). This method is only called if the last component is
576 negative or needs lookup. Cached positive dentries are still
577 handled by f_op->open(). If the file was created, FMODE_CREATED
578 flag should be set in file->f_mode. In case of O_EXCL the
579 method must only succeed if the file didn't exist and hence
580 FMODE_CREATED shall always be set on success.
583 called in the end of O_TMPFILE open(). Optional, equivalent to
584 atomically creating, opening and unlinking a file in given
588 The Address Space Object
589 ========================
591 The address space object is used to group and manage pages in the page
592 cache. It can be used to keep track of the pages in a file (or anything
593 else) and also track the mapping of sections of the file into process
596 There are a number of distinct yet related services that an
597 address-space can provide. These include communicating memory pressure,
598 page lookup by address, and keeping track of pages tagged as Dirty or
601 The first can be used independently to the others. The VM can try to
602 either write dirty pages in order to clean them, or release clean pages
603 in order to reuse them. To do this it can call the ->writepage method
604 on dirty pages, and ->releasepage on clean pages with PagePrivate set.
605 Clean pages without PagePrivate and with no external references will be
606 released without notice being given to the address_space.
608 To achieve this functionality, pages need to be placed on an LRU with
609 lru_cache_add and mark_page_active needs to be called whenever the page
612 Pages are normally kept in a radix tree index by ->index. This tree
613 maintains information about the PG_Dirty and PG_Writeback status of each
614 page, so that pages with either of these flags can be found quickly.
616 The Dirty tag is primarily used by mpage_writepages - the default
617 ->writepages method. It uses the tag to find dirty pages to call
618 ->writepage on. If mpage_writepages is not used (i.e. the address
619 provides its own ->writepages) , the PAGECACHE_TAG_DIRTY tag is almost
620 unused. write_inode_now and sync_inode do use it (through
621 __sync_single_inode) to check if ->writepages has been successful in
622 writing out the whole address_space.
624 The Writeback tag is used by filemap*wait* and sync_page* functions, via
625 filemap_fdatawait_range, to wait for all writeback to complete.
627 An address_space handler may attach extra information to a page,
628 typically using the 'private' field in the 'struct page'. If such
629 information is attached, the PG_Private flag should be set. This will
630 cause various VM routines to make extra calls into the address_space
631 handler to deal with that data.
633 An address space acts as an intermediate between storage and
634 application. Data is read into the address space a whole page at a
635 time, and provided to the application either by copying of the page, or
636 by memory-mapping the page. Data is written into the address space by
637 the application, and then written-back to storage typically in whole
638 pages, however the address_space has finer control of write sizes.
640 The read process essentially only requires 'readpage'. The write
641 process is more complicated and uses write_begin/write_end or
642 set_page_dirty to write data into the address_space, and writepage and
643 writepages to writeback data to storage.
645 Adding and removing pages to/from an address_space is protected by the
648 When data is written to a page, the PG_Dirty flag should be set. It
649 typically remains set until writepage asks for it to be written. This
650 should clear PG_Dirty and set PG_Writeback. It can be actually written
651 at any point after PG_Dirty is clear. Once it is known to be safe,
652 PG_Writeback is cleared.
654 Writeback makes use of a writeback_control structure to direct the
655 operations. This gives the the writepage and writepages operations some
656 information about the nature of and reason for the writeback request,
657 and the constraints under which it is being done. It is also used to
658 return information back to the caller about the result of a writepage or
662 Handling errors during writeback
663 --------------------------------
665 Most applications that do buffered I/O will periodically call a file
666 synchronization call (fsync, fdatasync, msync or sync_file_range) to
667 ensure that data written has made it to the backing store. When there
668 is an error during writeback, they expect that error to be reported when
669 a file sync request is made. After an error has been reported on one
670 request, subsequent requests on the same file descriptor should return
671 0, unless further writeback errors have occurred since the previous file
674 Ideally, the kernel would report errors only on file descriptions on
675 which writes were done that subsequently failed to be written back. The
676 generic pagecache infrastructure does not track the file descriptions
677 that have dirtied each individual page however, so determining which
678 file descriptors should get back an error is not possible.
680 Instead, the generic writeback error tracking infrastructure in the
681 kernel settles for reporting errors to fsync on all file descriptions
682 that were open at the time that the error occurred. In a situation with
683 multiple writers, all of them will get back an error on a subsequent
684 fsync, even if all of the writes done through that particular file
685 descriptor succeeded (or even if there were no writes on that file
688 Filesystems that wish to use this infrastructure should call
689 mapping_set_error to record the error in the address_space when it
690 occurs. Then, after writing back data from the pagecache in their
691 file->fsync operation, they should call file_check_and_advance_wb_err to
692 ensure that the struct file's error cursor has advanced to the correct
693 point in the stream of errors emitted by the backing device(s).
696 struct address_space_operations
697 -------------------------------
699 This describes how the VFS can manipulate mapping of a file to page
700 cache in your filesystem. The following members are defined:
704 struct address_space_operations {
705 int (*writepage)(struct page *page, struct writeback_control *wbc);
706 int (*readpage)(struct file *, struct page *);
707 int (*writepages)(struct address_space *, struct writeback_control *);
708 int (*set_page_dirty)(struct page *page);
709 int (*readpages)(struct file *filp, struct address_space *mapping,
710 struct list_head *pages, unsigned nr_pages);
711 int (*write_begin)(struct file *, struct address_space *mapping,
712 loff_t pos, unsigned len, unsigned flags,
713 struct page **pagep, void **fsdata);
714 int (*write_end)(struct file *, struct address_space *mapping,
715 loff_t pos, unsigned len, unsigned copied,
716 struct page *page, void *fsdata);
717 sector_t (*bmap)(struct address_space *, sector_t);
718 void (*invalidatepage) (struct page *, unsigned int, unsigned int);
719 int (*releasepage) (struct page *, int);
720 void (*freepage)(struct page *);
721 ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter);
722 /* isolate a page for migration */
723 bool (*isolate_page) (struct page *, isolate_mode_t);
724 /* migrate the contents of a page to the specified target */
725 int (*migratepage) (struct page *, struct page *);
726 /* put migration-failed page back to right list */
727 void (*putback_page) (struct page *);
728 int (*launder_page) (struct page *);
730 int (*is_partially_uptodate) (struct page *, unsigned long,
732 void (*is_dirty_writeback) (struct page *, bool *, bool *);
733 int (*error_remove_page) (struct mapping *mapping, struct page *page);
734 int (*swap_activate)(struct file *);
735 int (*swap_deactivate)(struct file *);
739 called by the VM to write a dirty page to backing store. This
740 may happen for data integrity reasons (i.e. 'sync'), or to free
741 up memory (flush). The difference can be seen in
742 wbc->sync_mode. The PG_Dirty flag has been cleared and
743 PageLocked is true. writepage should start writeout, should set
744 PG_Writeback, and should make sure the page is unlocked, either
745 synchronously or asynchronously when the write operation
748 If wbc->sync_mode is WB_SYNC_NONE, ->writepage doesn't have to
749 try too hard if there are problems, and may choose to write out
750 other pages from the mapping if that is easier (e.g. due to
751 internal dependencies). If it chooses not to start writeout, it
752 should return AOP_WRITEPAGE_ACTIVATE so that the VM will not
753 keep calling ->writepage on that page.
755 See the file "Locking" for more details.
758 called by the VM to read a page from backing store. The page
759 will be Locked when readpage is called, and should be unlocked
760 and marked uptodate once the read completes. If ->readpage
761 discovers that it needs to unlock the page for some reason, it
762 can do so, and then return AOP_TRUNCATED_PAGE. In this case,
763 the page will be relocated, relocked and if that all succeeds,
764 ->readpage will be called again.
767 called by the VM to write out pages associated with the
768 address_space object. If wbc->sync_mode is WBC_SYNC_ALL, then
769 the writeback_control will specify a range of pages that must be
770 written out. If it is WBC_SYNC_NONE, then a nr_to_write is
771 given and that many pages should be written if possible. If no
772 ->writepages is given, then mpage_writepages is used instead.
773 This will choose pages from the address space that are tagged as
774 DIRTY and will pass them to ->writepage.
777 called by the VM to set a page dirty. This is particularly
778 needed if an address space attaches private data to a page, and
779 that data needs to be updated when a page is dirtied. This is
780 called, for example, when a memory mapped page gets modified.
781 If defined, it should set the PageDirty flag, and the
782 PAGECACHE_TAG_DIRTY tag in the radix tree.
785 called by the VM to read pages associated with the address_space
786 object. This is essentially just a vector version of readpage.
787 Instead of just one page, several pages are requested.
788 readpages is only used for read-ahead, so read errors are
789 ignored. If anything goes wrong, feel free to give up.
792 Called by the generic buffered write code to ask the filesystem
793 to prepare to write len bytes at the given offset in the file.
794 The address_space should check that the write will be able to
795 complete, by allocating space if necessary and doing any other
796 internal housekeeping. If the write will update parts of any
797 basic-blocks on storage, then those blocks should be pre-read
798 (if they haven't been read already) so that the updated blocks
799 can be written out properly.
801 The filesystem must return the locked pagecache page for the
802 specified offset, in ``*pagep``, for the caller to write into.
804 It must be able to cope with short writes (where the length
805 passed to write_begin is greater than the number of bytes copied
808 flags is a field for AOP_FLAG_xxx flags, described in
811 A void * may be returned in fsdata, which then gets passed into
814 Returns 0 on success; < 0 on failure (which is the error code),
815 in which case write_end is not called.
818 After a successful write_begin, and data copy, write_end must be
819 called. len is the original len passed to write_begin, and
820 copied is the amount that was able to be copied.
822 The filesystem must take care of unlocking the page and
823 releasing it refcount, and updating i_size.
825 Returns < 0 on failure, otherwise the number of bytes (<=
826 'copied') that were able to be copied into pagecache.
829 called by the VFS to map a logical block offset within object to
830 physical block number. This method is used by the FIBMAP ioctl
831 and for working with swap-files. To be able to swap to a file,
832 the file must have a stable mapping to a block device. The swap
833 system does not go through the filesystem but instead uses bmap
834 to find out where the blocks in the file are and uses those
838 If a page has PagePrivate set, then invalidatepage will be
839 called when part or all of the page is to be removed from the
840 address space. This generally corresponds to either a
841 truncation, punch hole or a complete invalidation of the address
842 space (in the latter case 'offset' will always be 0 and 'length'
843 will be PAGE_SIZE). Any private data associated with the page
844 should be updated to reflect this truncation. If offset is 0
845 and length is PAGE_SIZE, then the private data should be
846 released, because the page must be able to be completely
847 discarded. This may be done by calling the ->releasepage
848 function, but in this case the release MUST succeed.
851 releasepage is called on PagePrivate pages to indicate that the
852 page should be freed if possible. ->releasepage should remove
853 any private data from the page and clear the PagePrivate flag.
854 If releasepage() fails for some reason, it must indicate failure
855 with a 0 return value. releasepage() is used in two distinct
856 though related cases. The first is when the VM finds a clean
857 page with no active users and wants to make it a free page. If
858 ->releasepage succeeds, the page will be removed from the
859 address_space and become free.
861 The second case is when a request has been made to invalidate
862 some or all pages in an address_space. This can happen through
863 the fadvise(POSIX_FADV_DONTNEED) system call or by the
864 filesystem explicitly requesting it as nfs and 9fs do (when they
865 believe the cache may be out of date with storage) by calling
866 invalidate_inode_pages2(). If the filesystem makes such a call,
867 and needs to be certain that all pages are invalidated, then its
868 releasepage will need to ensure this. Possibly it can clear the
869 PageUptodate bit if it cannot free private data yet.
872 freepage is called once the page is no longer visible in the
873 page cache in order to allow the cleanup of any private data.
874 Since it may be called by the memory reclaimer, it should not
875 assume that the original address_space mapping still exists, and
879 called by the generic read/write routines to perform direct_IO -
880 that is IO requests which bypass the page cache and transfer
881 data directly between the storage and the application's address
885 Called by the VM when isolating a movable non-lru page. If page
886 is successfully isolated, VM marks the page as PG_isolated via
890 This is used to compact the physical memory usage. If the VM
891 wants to relocate a page (maybe off a memory card that is
892 signalling imminent failure) it will pass a new page and an old
893 page to this function. migrate_page should transfer any private
894 data across and update any references that it has to the page.
897 Called by the VM when isolated page's migration fails.
900 Called before freeing a page - it writes back the dirty page.
901 To prevent redirtying the page, it is kept locked during the
904 ``is_partially_uptodate``
905 Called by the VM when reading a file through the pagecache when
906 the underlying blocksize != pagesize. If the required block is
907 up to date then the read can complete without needing the IO to
908 bring the whole page up to date.
910 ``is_dirty_writeback``
911 Called by the VM when attempting to reclaim a page. The VM uses
912 dirty and writeback information to determine if it needs to
913 stall to allow flushers a chance to complete some IO.
914 Ordinarily it can use PageDirty and PageWriteback but some
915 filesystems have more complex state (unstable pages in NFS
916 prevent reclaim) or do not set those flags due to locking
917 problems. This callback allows a filesystem to indicate to the
918 VM if a page should be treated as dirty or writeback for the
919 purposes of stalling.
921 ``error_remove_page``
922 normally set to generic_error_remove_page if truncation is ok
923 for this address space. Used for memory failure handling.
924 Setting this implies you deal with pages going away under you,
925 unless you have them locked or reference counts increased.
928 Called when swapon is used on a file to allocate space if
929 necessary and pin the block lookup information in memory. A
930 return value of zero indicates success, in which case this file
931 can be used to back swapspace.
934 Called during swapoff on files where swap_activate was
941 A file object represents a file opened by a process. This is also known
942 as an "open file description" in POSIX parlance.
945 struct file_operations
946 ----------------------
948 This describes how the VFS can manipulate an open file. As of kernel
949 4.18, the following members are defined:
953 struct file_operations {
954 struct module *owner;
955 loff_t (*llseek) (struct file *, loff_t, int);
956 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
957 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
958 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
959 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
960 int (*iopoll)(struct kiocb *kiocb, bool spin);
961 int (*iterate) (struct file *, struct dir_context *);
962 int (*iterate_shared) (struct file *, struct dir_context *);
963 __poll_t (*poll) (struct file *, struct poll_table_struct *);
964 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
965 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
966 int (*mmap) (struct file *, struct vm_area_struct *);
967 int (*open) (struct inode *, struct file *);
968 int (*flush) (struct file *, fl_owner_t id);
969 int (*release) (struct inode *, struct file *);
970 int (*fsync) (struct file *, loff_t, loff_t, int datasync);
971 int (*fasync) (int, struct file *, int);
972 int (*lock) (struct file *, int, struct file_lock *);
973 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
974 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
975 int (*check_flags)(int);
976 int (*flock) (struct file *, int, struct file_lock *);
977 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
978 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
979 int (*setlease)(struct file *, long, struct file_lock **, void **);
980 long (*fallocate)(struct file *file, int mode, loff_t offset,
982 void (*show_fdinfo)(struct seq_file *m, struct file *f);
984 unsigned (*mmap_capabilities)(struct file *);
986 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *, loff_t, size_t, unsigned int);
987 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in,
988 struct file *file_out, loff_t pos_out,
989 loff_t len, unsigned int remap_flags);
990 int (*fadvise)(struct file *, loff_t, loff_t, int);
993 Again, all methods are called without any locks being held, unless
997 called when the VFS needs to move the file position index
1000 called by read(2) and related system calls
1003 possibly asynchronous read with iov_iter as destination
1006 called by write(2) and related system calls
1009 possibly asynchronous write with iov_iter as source
1012 called when aio wants to poll for completions on HIPRI iocbs
1015 called when the VFS needs to read the directory contents
1018 called when the VFS needs to read the directory contents when
1019 filesystem supports concurrent dir iterators
1022 called by the VFS when a process wants to check if there is
1023 activity on this file and (optionally) go to sleep until there
1024 is activity. Called by the select(2) and poll(2) system calls
1027 called by the ioctl(2) system call.
1030 called by the ioctl(2) system call when 32 bit system calls are
1031 used on 64 bit kernels.
1034 called by the mmap(2) system call
1037 called by the VFS when an inode should be opened. When the VFS
1038 opens a file, it creates a new "struct file". It then calls the
1039 open method for the newly allocated file structure. You might
1040 think that the open method really belongs in "struct
1041 inode_operations", and you may be right. I think it's done the
1042 way it is because it makes filesystems simpler to implement.
1043 The open() method is a good place to initialize the
1044 "private_data" member in the file structure if you want to point
1045 to a device structure
1048 called by the close(2) system call to flush a file
1051 called when the last reference to an open file is closed
1054 called by the fsync(2) system call. Also see the section above
1055 entitled "Handling errors during writeback".
1058 called by the fcntl(2) system call when asynchronous
1059 (non-blocking) mode is enabled for a file
1062 called by the fcntl(2) system call for F_GETLK, F_SETLK, and
1065 ``get_unmapped_area``
1066 called by the mmap(2) system call
1069 called by the fcntl(2) system call for F_SETFL command
1072 called by the flock(2) system call
1075 called by the VFS to splice data from a pipe to a file. This
1076 method is used by the splice(2) system call
1079 called by the VFS to splice data from file to a pipe. This
1080 method is used by the splice(2) system call
1083 called by the VFS to set or release a file lock lease. setlease
1084 implementations should call generic_setlease to record or remove
1085 the lease in the inode after setting it.
1088 called by the VFS to preallocate blocks or punch a hole.
1091 called by the copy_file_range(2) system call.
1093 ``remap_file_range``
1094 called by the ioctl(2) system call for FICLONERANGE and FICLONE
1095 and FIDEDUPERANGE commands to remap file ranges. An
1096 implementation should remap len bytes at pos_in of the source
1097 file into the dest file at pos_out. Implementations must handle
1098 callers passing in len == 0; this means "remap to the end of the
1099 source file". The return value should the number of bytes
1100 remapped, or the usual negative error code if errors occurred
1101 before any bytes were remapped. The remap_flags parameter
1102 accepts REMAP_FILE_* flags. If REMAP_FILE_DEDUP is set then the
1103 implementation must only remap if the requested file ranges have
1104 identical contents. If REMAP_CAN_SHORTEN is set, the caller is
1105 ok with the implementation shortening the request length to
1106 satisfy alignment or EOF requirements (or any other reason).
1109 possibly called by the fadvise64() system call.
1111 Note that the file operations are implemented by the specific
1112 filesystem in which the inode resides. When opening a device node
1113 (character or block special) most filesystems will call special
1114 support routines in the VFS which will locate the required device
1115 driver information. These support routines replace the filesystem file
1116 operations with those for the device driver, and then proceed to call
1117 the new open() method for the file. This is how opening a device file
1118 in the filesystem eventually ends up calling the device driver open()
1122 Directory Entry Cache (dcache)
1123 ==============================
1126 struct dentry_operations
1127 ------------------------
1129 This describes how a filesystem can overload the standard dentry
1130 operations. Dentries and the dcache are the domain of the VFS and the
1131 individual filesystem implementations. Device drivers have no business
1132 here. These methods may be set to NULL, as they are either optional or
1133 the VFS uses a default. As of kernel 2.6.22, the following members are
1138 struct dentry_operations {
1139 int (*d_revalidate)(struct dentry *, unsigned int);
1140 int (*d_weak_revalidate)(struct dentry *, unsigned int);
1141 int (*d_hash)(const struct dentry *, struct qstr *);
1142 int (*d_compare)(const struct dentry *,
1143 unsigned int, const char *, const struct qstr *);
1144 int (*d_delete)(const struct dentry *);
1145 int (*d_init)(struct dentry *);
1146 void (*d_release)(struct dentry *);
1147 void (*d_iput)(struct dentry *, struct inode *);
1148 char *(*d_dname)(struct dentry *, char *, int);
1149 struct vfsmount *(*d_automount)(struct path *);
1150 int (*d_manage)(const struct path *, bool);
1151 struct dentry *(*d_real)(struct dentry *, const struct inode *);
1155 called when the VFS needs to revalidate a dentry. This is
1156 called whenever a name look-up finds a dentry in the dcache.
1157 Most local filesystems leave this as NULL, because all their
1158 dentries in the dcache are valid. Network filesystems are
1159 different since things can change on the server without the
1160 client necessarily being aware of it.
1162 This function should return a positive value if the dentry is
1163 still valid, and zero or a negative error code if it isn't.
1165 d_revalidate may be called in rcu-walk mode (flags &
1166 LOOKUP_RCU). If in rcu-walk mode, the filesystem must
1167 revalidate the dentry without blocking or storing to the dentry,
1168 d_parent and d_inode should not be used without care (because
1169 they can change and, in d_inode case, even become NULL under
1172 If a situation is encountered that rcu-walk cannot handle,
1174 -ECHILD and it will be called again in ref-walk mode.
1176 ``_weak_revalidate``
1177 called when the VFS needs to revalidate a "jumped" dentry. This
1178 is called when a path-walk ends at dentry that was not acquired
1179 by doing a lookup in the parent directory. This includes "/",
1180 "." and "..", as well as procfs-style symlinks and mountpoint
1183 In this case, we are less concerned with whether the dentry is
1184 still fully correct, but rather that the inode is still valid.
1185 As with d_revalidate, most local filesystems will set this to
1186 NULL since their dcache entries are always valid.
1188 This function has the same return code semantics as
1191 d_weak_revalidate is only called after leaving rcu-walk mode.
1194 called when the VFS adds a dentry to the hash table. The first
1195 dentry passed to d_hash is the parent directory that the name is
1198 Same locking and synchronisation rules as d_compare regarding
1199 what is safe to dereference etc.
1202 called to compare a dentry name with a given name. The first
1203 dentry is the parent of the dentry to be compared, the second is
1204 the child dentry. len and name string are properties of the
1205 dentry to be compared. qstr is the name to compare it with.
1207 Must be constant and idempotent, and should not take locks if
1208 possible, and should not or store into the dentry. Should not
1209 dereference pointers outside the dentry without lots of care
1210 (eg. d_parent, d_inode, d_name should not be used).
1212 However, our vfsmount is pinned, and RCU held, so the dentries
1213 and inodes won't disappear, neither will our sb or filesystem
1214 module. ->d_sb may be used.
1216 It is a tricky calling convention because it needs to be called
1217 under "rcu-walk", ie. without any locks or references on things.
1220 called when the last reference to a dentry is dropped and the
1221 dcache is deciding whether or not to cache it. Return 1 to
1222 delete immediately, or 0 to cache the dentry. Default is NULL
1223 which means to always cache a reachable dentry. d_delete must
1224 be constant and idempotent.
1227 called when a dentry is allocated
1230 called when a dentry is really deallocated
1233 called when a dentry loses its inode (just prior to its being
1234 deallocated). The default when this is NULL is that the VFS
1235 calls iput(). If you define this method, you must call iput()
1239 called when the pathname of a dentry should be generated.
1240 Useful for some pseudo filesystems (sockfs, pipefs, ...) to
1241 delay pathname generation. (Instead of doing it when dentry is
1242 created, it's done only when the path is needed.). Real
1243 filesystems probably dont want to use it, because their dentries
1244 are present in global dcache hash, so their hash should be an
1245 invariant. As no lock is held, d_dname() should not try to
1246 modify the dentry itself, unless appropriate SMP safety is used.
1247 CAUTION : d_path() logic is quite tricky. The correct way to
1248 return for example "Hello" is to put it at the end of the
1249 buffer, and returns a pointer to the first char.
1250 dynamic_dname() helper function is provided to take care of
1257 static char *pipefs_dname(struct dentry *dent, char *buffer, int buflen)
1259 return dynamic_dname(dentry, buffer, buflen, "pipe:[%lu]",
1260 dentry->d_inode->i_ino);
1264 called when an automount dentry is to be traversed (optional).
1265 This should create a new VFS mount record and return the record
1266 to the caller. The caller is supplied with a path parameter
1267 giving the automount directory to describe the automount target
1268 and the parent VFS mount record to provide inheritable mount
1269 parameters. NULL should be returned if someone else managed to
1270 make the automount first. If the vfsmount creation failed, then
1271 an error code should be returned. If -EISDIR is returned, then
1272 the directory will be treated as an ordinary directory and
1273 returned to pathwalk to continue walking.
1275 If a vfsmount is returned, the caller will attempt to mount it
1276 on the mountpoint and will remove the vfsmount from its
1277 expiration list in the case of failure. The vfsmount should be
1278 returned with 2 refs on it to prevent automatic expiration - the
1279 caller will clean up the additional ref.
1281 This function is only used if DCACHE_NEED_AUTOMOUNT is set on
1282 the dentry. This is set by __d_instantiate() if S_AUTOMOUNT is
1283 set on the inode being added.
1286 called to allow the filesystem to manage the transition from a
1287 dentry (optional). This allows autofs, for example, to hold up
1288 clients waiting to explore behind a 'mountpoint' while letting
1289 the daemon go past and construct the subtree there. 0 should be
1290 returned to let the calling process continue. -EISDIR can be
1291 returned to tell pathwalk to use this directory as an ordinary
1292 directory and to ignore anything mounted on it and not to check
1293 the automount flag. Any other error code will abort pathwalk
1296 If the 'rcu_walk' parameter is true, then the caller is doing a
1297 pathwalk in RCU-walk mode. Sleeping is not permitted in this
1298 mode, and the caller can be asked to leave it and call again by
1299 returning -ECHILD. -EISDIR may also be returned to tell
1300 pathwalk to ignore d_automount or any mounts.
1302 This function is only used if DCACHE_MANAGE_TRANSIT is set on
1303 the dentry being transited from.
1306 overlay/union type filesystems implement this method to return
1307 one of the underlying dentries hidden by the overlay. It is
1308 used in two different modes:
1310 Called from file_dentry() it returns the real dentry matching
1311 the inode argument. The real dentry may be from a lower layer
1312 already copied up, but still referenced from the file. This
1313 mode is selected with a non-NULL inode argument.
1315 With NULL inode the topmost real underlying dentry is returned.
1317 Each dentry has a pointer to its parent dentry, as well as a hash list
1318 of child dentries. Child dentries are basically like files in a
1322 Directory Entry Cache API
1323 --------------------------
1325 There are a number of functions defined which permit a filesystem to
1326 manipulate dentries:
1329 open a new handle for an existing dentry (this just increments
1333 close a handle for a dentry (decrements the usage count). If
1334 the usage count drops to 0, and the dentry is still in its
1335 parent's hash, the "d_delete" method is called to check whether
1336 it should be cached. If it should not be cached, or if the
1337 dentry is not hashed, it is deleted. Otherwise cached dentries
1338 are put into an LRU list to be reclaimed on memory shortage.
1341 this unhashes a dentry from its parents hash list. A subsequent
1342 call to dput() will deallocate the dentry if its usage count
1346 delete a dentry. If there are no other open references to the
1347 dentry then the dentry is turned into a negative dentry (the
1348 d_iput() method is called). If there are other references, then
1349 d_drop() is called instead
1352 add a dentry to its parents hash list and then calls
1356 add a dentry to the alias hash list for the inode and updates
1357 the "d_inode" member. The "i_count" member in the inode
1358 structure should be set/incremented. If the inode pointer is
1359 NULL, the dentry is called a "negative dentry". This function
1360 is commonly called when an inode is created for an existing
1364 look up a dentry given its parent and path name component It
1365 looks up the child of that given name from the dcache hash
1366 table. If it is found, the reference count is incremented and
1367 the dentry is returned. The caller must use dput() to free the
1368 dentry when it finishes using it.
1378 On mount and remount the filesystem is passed a string containing a
1379 comma separated list of mount options. The options can have either of
1385 The <linux/parser.h> header defines an API that helps parse these
1386 options. There are plenty of examples on how to use it in existing
1393 If a filesystem accepts mount options, it must define show_options() to
1394 show all the currently active options. The rules are:
1396 - options MUST be shown which are not default or their values differ
1399 - options MAY be shown which are enabled by default or have their
1402 Options used only internally between a mount helper and the kernel (such
1403 as file descriptors), or which only have an effect during the mounting
1404 (such as ones controlling the creation of a journal) are exempt from the
1407 The underlying reason for the above rules is to make sure, that a mount
1408 can be accurately replicated (e.g. umounting and mounting again) based
1409 on the information found in /proc/mounts.
1415 (Note some of these resources are not up-to-date with the latest kernel
1418 Creating Linux virtual filesystems. 2002
1419 <http://lwn.net/Articles/13325/>
1421 The Linux Virtual File-system Layer by Neil Brown. 1999
1422 <http://www.cse.unsw.edu.au/~neilb/oss/linux-commentary/vfs.html>
1424 A tour of the Linux VFS by Michael K. Johnson. 1996
1425 <http://www.tldp.org/LDP/khg/HyperNews/get/fs/vfstour.html>
1427 A small trail through the Linux kernel by Andries Brouwer. 2001
1428 <http://www.win.tue.nl/~aeb/linux/vfs/trail.html>
projects / mirror_ubuntu-hirsute-kernel.git / blob