2 Overview of the Linux Virtual File System
4 Original author: Richard Gooch <rgooch@atnf.csiro.au>
6 Last updated on June 24, 2007.
8 Copyright (C) 1999 Richard Gooch
9 Copyright (C) 2005 Pekka Enberg
11 This file is released under the GPLv2.
17 The Virtual File System (also known as the Virtual Filesystem Switch)
18 is the software layer in the kernel that provides the filesystem
19 interface to userspace programs. It also provides an abstraction
20 within the kernel which allows different filesystem implementations to
23 VFS system calls open(2), stat(2), read(2), write(2), chmod(2) and so
24 on are called from a process context. Filesystem locking is described
25 in the document Documentation/filesystems/Locking.
28 Directory Entry Cache (dcache)
29 ------------------------------
31 The VFS implements the open(2), stat(2), chmod(2), and similar system
32 calls. The pathname argument that is passed to them is used by the VFS
33 to search through the directory entry cache (also known as the dentry
34 cache or dcache). This provides a very fast look-up mechanism to
35 translate a pathname (filename) into a specific dentry. Dentries live
36 in RAM and are never saved to disc: they exist only for performance.
38 The dentry cache is meant to be a view into your entire filespace. As
39 most computers cannot fit all dentries in the RAM at the same time,
40 some bits of the cache are missing. In order to resolve your pathname
41 into a dentry, the VFS may have to resort to creating dentries along
42 the way, and then loading the inode. This is done by looking up the
49 An individual dentry usually has a pointer to an inode. Inodes are
50 filesystem objects such as regular files, directories, FIFOs and other
51 beasts. They live either on the disc (for block device filesystems)
52 or in the memory (for pseudo filesystems). Inodes that live on the
53 disc are copied into the memory when required and changes to the inode
54 are written back to disc. A single inode can be pointed to by multiple
55 dentries (hard links, for example, do this).
57 To look up an inode requires that the VFS calls the lookup() method of
58 the parent directory inode. This method is installed by the specific
59 filesystem implementation that the inode lives in. Once the VFS has
60 the required dentry (and hence the inode), we can do all those boring
61 things like open(2) the file, or stat(2) it to peek at the inode
62 data. The stat(2) operation is fairly simple: once the VFS has the
63 dentry, it peeks at the inode data and passes some of it back to
70 Opening a file requires another operation: allocation of a file
71 structure (this is the kernel-side implementation of file
72 descriptors). The freshly allocated file structure is initialized with
73 a pointer to the dentry and a set of file operation member functions.
74 These are taken from the inode data. The open() file method is then
75 called so the specific filesystem implementation can do its work. You
76 can see that this is another switch performed by the VFS. The file
77 structure is placed into the file descriptor table for the process.
79 Reading, writing and closing files (and other assorted VFS operations)
80 is done by using the userspace file descriptor to grab the appropriate
81 file structure, and then calling the required file structure method to
82 do whatever is required. For as long as the file is open, it keeps the
83 dentry in use, which in turn means that the VFS inode is still in use.
86 Registering and Mounting a Filesystem
87 =====================================
89 To register and unregister a filesystem, use the following API
94 extern int register_filesystem(struct file_system_type *);
95 extern int unregister_filesystem(struct file_system_type *);
97 The passed struct file_system_type describes your filesystem. When a
98 request is made to mount a device onto a directory in your filespace,
99 the VFS will call the appropriate get_sb() method for the specific
100 filesystem. The dentry for the mount point will then be updated to
101 point to the root inode for the new filesystem.
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 2.6.22, the following
113 struct file_system_type {
116 int (*get_sb) (struct file_system_type *, int,
117 const char *, void *, struct vfsmount *);
118 void (*kill_sb) (struct super_block *);
119 struct module *owner;
120 struct file_system_type * next;
121 struct list_head fs_supers;
122 struct lock_class_key s_lock_key;
123 struct lock_class_key s_umount_key;
126 name: the name of the filesystem type, such as "ext2", "iso9660",
129 fs_flags: various flags (i.e. FS_REQUIRES_DEV, FS_NO_DCACHE, etc.)
131 get_sb: the method to call when a new instance of this
132 filesystem should be mounted
134 kill_sb: the method to call when an instance of this filesystem
137 owner: for internal VFS use: you should initialize this to THIS_MODULE in
140 next: for internal VFS use: you should initialize this to NULL
142 s_lock_key, s_umount_key: lockdep-specific
144 The get_sb() method has the following arguments:
146 struct file_system_type *fs_type: describes the filesystem, partly initialized
147 by the specific filesystem code
149 int flags: mount flags
151 const char *dev_name: the device name we are mounting.
153 void *data: arbitrary mount options, usually comes as an ASCII
154 string (see "Mount Options" section)
156 struct vfsmount *mnt: a vfs-internal representation of a mount point
158 The get_sb() method must determine if the block device specified
159 in the dev_name and fs_type contains a filesystem of the type the method
160 supports. If it succeeds in opening the named block device, it initializes a
161 struct super_block descriptor for the filesystem contained by the block device.
162 On failure it returns an error.
164 The most interesting member of the superblock structure that the
165 get_sb() method fills in is the "s_op" field. This is a pointer to
166 a "struct super_operations" which describes the next level of the
167 filesystem implementation.
169 Usually, a filesystem uses one of the generic get_sb() implementations
170 and provides a fill_super() method instead. The generic methods are:
172 get_sb_bdev: mount a filesystem residing on a block device
174 get_sb_nodev: mount a filesystem that is not backed by a device
176 get_sb_single: mount a filesystem which shares the instance between
179 A fill_super() method implementation has the following arguments:
181 struct super_block *sb: the superblock structure. The method fill_super()
182 must initialize this properly.
184 void *data: arbitrary mount options, usually comes as an ASCII
185 string (see "Mount Options" section)
187 int silent: whether or not to be silent on error
190 The Superblock Object
191 =====================
193 A superblock object represents a mounted filesystem.
196 struct super_operations
197 -----------------------
199 This describes how the VFS can manipulate the superblock of your
200 filesystem. As of kernel 2.6.22, the following members are defined:
202 struct super_operations {
203 struct inode *(*alloc_inode)(struct super_block *sb);
204 void (*destroy_inode)(struct inode *);
206 void (*dirty_inode) (struct inode *);
207 int (*write_inode) (struct inode *, int);
208 void (*drop_inode) (struct inode *);
209 void (*delete_inode) (struct inode *);
210 void (*put_super) (struct super_block *);
211 void (*write_super) (struct super_block *);
212 int (*sync_fs)(struct super_block *sb, int wait);
213 int (*freeze_fs) (struct super_block *);
214 int (*unfreeze_fs) (struct super_block *);
215 int (*statfs) (struct dentry *, struct kstatfs *);
216 int (*remount_fs) (struct super_block *, int *, char *);
217 void (*clear_inode) (struct inode *);
218 void (*umount_begin) (struct super_block *);
220 int (*show_options)(struct seq_file *, struct vfsmount *);
222 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
223 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
226 All methods are called without any locks being held, unless otherwise
227 noted. This means that most methods can block safely. All methods are
228 only called from a process context (i.e. not from an interrupt handler
231 alloc_inode: this method is called by inode_alloc() to allocate memory
232 for struct inode and initialize it. If this function is not
233 defined, a simple 'struct inode' is allocated. Normally
234 alloc_inode will be used to allocate a larger structure which
235 contains a 'struct inode' embedded within it.
237 destroy_inode: this method is called by destroy_inode() to release
238 resources allocated for struct inode. It is only required if
239 ->alloc_inode was defined and simply undoes anything done by
242 dirty_inode: this method is called by the VFS to mark an inode dirty.
244 write_inode: this method is called when the VFS needs to write an
245 inode to disc. The second parameter indicates whether the write
246 should be synchronous or not, not all filesystems check this flag.
248 drop_inode: called when the last access to the inode is dropped,
249 with the inode_lock spinlock held.
251 This method should be either NULL (normal UNIX filesystem
252 semantics) or "generic_delete_inode" (for filesystems that do not
253 want to cache inodes - causing "delete_inode" to always be
254 called regardless of the value of i_nlink)
256 The "generic_delete_inode()" behavior is equivalent to the
257 old practice of using "force_delete" in the put_inode() case,
258 but does not have the races that the "force_delete()" approach
261 delete_inode: called when the VFS wants to delete an inode
263 put_super: called when the VFS wishes to free the superblock
264 (i.e. unmount). This is called with the superblock lock held
266 write_super: called when the VFS superblock needs to be written to
267 disc. This method is optional
269 sync_fs: called when VFS is writing out all dirty data associated with
270 a superblock. The second parameter indicates whether the method
271 should wait until the write out has been completed. Optional.
273 freeze_fs: called when VFS is locking a filesystem and
274 forcing it into a consistent state. This method is currently
275 used by the Logical Volume Manager (LVM).
277 unfreeze_fs: called when VFS is unlocking a filesystem and making it writable
280 statfs: called when the VFS needs to get filesystem statistics.
282 remount_fs: called when the filesystem is remounted. This is called
283 with the kernel lock held
285 clear_inode: called then the VFS clears the inode. Optional
287 umount_begin: called when the VFS is unmounting a filesystem.
289 show_options: called by the VFS to show mount options for
290 /proc/<pid>/mounts. (see "Mount Options" section)
292 quota_read: called by the VFS to read from filesystem quota file.
294 quota_write: called by the VFS to write to filesystem quota file.
296 Whoever sets up the inode is responsible for filling in the "i_op" field. This
297 is a pointer to a "struct inode_operations" which describes the methods that
298 can be performed on individual inodes.
304 An inode object represents an object within the filesystem.
307 struct inode_operations
308 -----------------------
310 This describes how the VFS can manipulate an inode in your
311 filesystem. As of kernel 2.6.22, the following members are defined:
313 struct inode_operations {
314 int (*create) (struct inode *,struct dentry *,int, struct nameidata *);
315 struct dentry * (*lookup) (struct inode *,struct dentry *, struct nameidata *);
316 int (*link) (struct dentry *,struct inode *,struct dentry *);
317 int (*unlink) (struct inode *,struct dentry *);
318 int (*symlink) (struct inode *,struct dentry *,const char *);
319 int (*mkdir) (struct inode *,struct dentry *,int);
320 int (*rmdir) (struct inode *,struct dentry *);
321 int (*mknod) (struct inode *,struct dentry *,int,dev_t);
322 int (*rename) (struct inode *, struct dentry *,
323 struct inode *, struct dentry *);
324 int (*readlink) (struct dentry *, char __user *,int);
325 void * (*follow_link) (struct dentry *, struct nameidata *);
326 void (*put_link) (struct dentry *, struct nameidata *, void *);
327 void (*truncate) (struct inode *);
328 int (*permission) (struct inode *, int, struct nameidata *);
329 int (*setattr) (struct dentry *, struct iattr *);
330 int (*getattr) (struct vfsmount *mnt, struct dentry *, struct kstat *);
331 int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
332 ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t);
333 ssize_t (*listxattr) (struct dentry *, char *, size_t);
334 int (*removexattr) (struct dentry *, const char *);
335 void (*truncate_range)(struct inode *, loff_t, loff_t);
338 Again, all methods are called without any locks being held, unless
341 create: called by the open(2) and creat(2) system calls. Only
342 required if you want to support regular files. The dentry you
343 get should not have an inode (i.e. it should be a negative
344 dentry). Here you will probably call d_instantiate() with the
345 dentry and the newly created inode
347 lookup: called when the VFS needs to look up an inode in a parent
348 directory. The name to look for is found in the dentry. This
349 method must call d_add() to insert the found inode into the
350 dentry. The "i_count" field in the inode structure should be
351 incremented. If the named inode does not exist a NULL inode
352 should be inserted into the dentry (this is called a negative
353 dentry). Returning an error code from this routine must only
354 be done on a real error, otherwise creating inodes with system
355 calls like create(2), mknod(2), mkdir(2) and so on will fail.
356 If you wish to overload the dentry methods then you should
357 initialise the "d_dop" field in the dentry; this is a pointer
358 to a struct "dentry_operations".
359 This method is called with the directory inode semaphore held
361 link: called by the link(2) system call. Only required if you want
362 to support hard links. You will probably need to call
363 d_instantiate() just as you would in the create() method
365 unlink: called by the unlink(2) system call. Only required if you
366 want to support deleting inodes
368 symlink: called by the symlink(2) system call. Only required if you
369 want to support symlinks. You will probably need to call
370 d_instantiate() just as you would in the create() method
372 mkdir: called by the mkdir(2) system call. Only required if you want
373 to support creating subdirectories. You will probably need to
374 call d_instantiate() just as you would in the create() method
376 rmdir: called by the rmdir(2) system call. Only required if you want
377 to support deleting subdirectories
379 mknod: called by the mknod(2) system call to create a device (char,
380 block) inode or a named pipe (FIFO) or socket. Only required
381 if you want to support creating these types of inodes. You
382 will probably need to call d_instantiate() just as you would
383 in the create() method
385 rename: called by the rename(2) system call to rename the object to
386 have the parent and name given by the second inode and dentry.
388 readlink: called by the readlink(2) system call. Only required if
389 you want to support reading symbolic links
391 follow_link: called by the VFS to follow a symbolic link to the
392 inode it points to. Only required if you want to support
393 symbolic links. This method returns a void pointer cookie
394 that is passed to put_link().
396 put_link: called by the VFS to release resources allocated by
397 follow_link(). The cookie returned by follow_link() is passed
398 to this method as the last parameter. It is used by
399 filesystems such as NFS where page cache is not stable
400 (i.e. page that was installed when the symbolic link walk
401 started might not be in the page cache at the end of the
404 truncate: Deprecated. This will not be called if ->setsize is defined.
405 Called by the VFS to change the size of a file. The
406 i_size field of the inode is set to the desired size by the
407 VFS before this method is called. This method is called by
408 the truncate(2) system call and related functionality.
410 Note: ->truncate and vmtruncate are deprecated. Do not add new
411 instances/calls of these. Filesystems should be converted to do their
412 truncate sequence via ->setattr().
414 permission: called by the VFS to check for access rights on a POSIX-like
417 setattr: called by the VFS to set attributes for a file. This method
418 is called by chmod(2) and related system calls.
420 getattr: called by the VFS to get attributes of a file. This method
421 is called by stat(2) and related system calls.
423 setxattr: called by the VFS to set an extended attribute for a file.
424 Extended attribute is a name:value pair associated with an
425 inode. This method is called by setxattr(2) system call.
427 getxattr: called by the VFS to retrieve the value of an extended
428 attribute name. This method is called by getxattr(2) function
431 listxattr: called by the VFS to list all extended attributes for a
432 given file. This method is called by listxattr(2) system call.
434 removexattr: called by the VFS to remove an extended attribute from
435 a file. This method is called by removexattr(2) system call.
437 truncate_range: a method provided by the underlying filesystem to truncate a
438 range of blocks , i.e. punch a hole somewhere in a file.
441 The Address Space Object
442 ========================
444 The address space object is used to group and manage pages in the page
445 cache. It can be used to keep track of the pages in a file (or
446 anything else) and also track the mapping of sections of the file into
447 process address spaces.
449 There are a number of distinct yet related services that an
450 address-space can provide. These include communicating memory
451 pressure, page lookup by address, and keeping track of pages tagged as
454 The first can be used independently to the others. The VM can try to
455 either write dirty pages in order to clean them, or release clean
456 pages in order to reuse them. To do this it can call the ->writepage
457 method on dirty pages, and ->releasepage on clean pages with
458 PagePrivate set. Clean pages without PagePrivate and with no external
459 references will be released without notice being given to the
462 To achieve this functionality, pages need to be placed on an LRU with
463 lru_cache_add and mark_page_active needs to be called whenever the
466 Pages are normally kept in a radix tree index by ->index. This tree
467 maintains information about the PG_Dirty and PG_Writeback status of
468 each page, so that pages with either of these flags can be found
471 The Dirty tag is primarily used by mpage_writepages - the default
472 ->writepages method. It uses the tag to find dirty pages to call
473 ->writepage on. If mpage_writepages is not used (i.e. the address
474 provides its own ->writepages) , the PAGECACHE_TAG_DIRTY tag is
475 almost unused. write_inode_now and sync_inode do use it (through
476 __sync_single_inode) to check if ->writepages has been successful in
477 writing out the whole address_space.
479 The Writeback tag is used by filemap*wait* and sync_page* functions,
480 via filemap_fdatawait_range, to wait for all writeback to
481 complete. While waiting ->sync_page (if defined) will be called on
482 each page that is found to require writeback.
484 An address_space handler may attach extra information to a page,
485 typically using the 'private' field in the 'struct page'. If such
486 information is attached, the PG_Private flag should be set. This will
487 cause various VM routines to make extra calls into the address_space
488 handler to deal with that data.
490 An address space acts as an intermediate between storage and
491 application. Data is read into the address space a whole page at a
492 time, and provided to the application either by copying of the page,
493 or by memory-mapping the page.
494 Data is written into the address space by the application, and then
495 written-back to storage typically in whole pages, however the
496 address_space has finer control of write sizes.
498 The read process essentially only requires 'readpage'. The write
499 process is more complicated and uses write_begin/write_end or
500 set_page_dirty to write data into the address_space, and writepage,
501 sync_page, and writepages to writeback data to storage.
503 Adding and removing pages to/from an address_space is protected by the
506 When data is written to a page, the PG_Dirty flag should be set. It
507 typically remains set until writepage asks for it to be written. This
508 should clear PG_Dirty and set PG_Writeback. It can be actually
509 written at any point after PG_Dirty is clear. Once it is known to be
510 safe, PG_Writeback is cleared.
512 Writeback makes use of a writeback_control structure...
514 struct address_space_operations
515 -------------------------------
517 This describes how the VFS can manipulate mapping of a file to page cache in
518 your filesystem. As of kernel 2.6.22, the following members are defined:
520 struct address_space_operations {
521 int (*writepage)(struct page *page, struct writeback_control *wbc);
522 int (*readpage)(struct file *, struct page *);
523 int (*sync_page)(struct page *);
524 int (*writepages)(struct address_space *, struct writeback_control *);
525 int (*set_page_dirty)(struct page *page);
526 int (*readpages)(struct file *filp, struct address_space *mapping,
527 struct list_head *pages, unsigned nr_pages);
528 int (*write_begin)(struct file *, struct address_space *mapping,
529 loff_t pos, unsigned len, unsigned flags,
530 struct page **pagep, void **fsdata);
531 int (*write_end)(struct file *, struct address_space *mapping,
532 loff_t pos, unsigned len, unsigned copied,
533 struct page *page, void *fsdata);
534 sector_t (*bmap)(struct address_space *, sector_t);
535 int (*invalidatepage) (struct page *, unsigned long);
536 int (*releasepage) (struct page *, int);
537 void (*freepage)(struct page *);
538 ssize_t (*direct_IO)(int, struct kiocb *, const struct iovec *iov,
539 loff_t offset, unsigned long nr_segs);
540 struct page* (*get_xip_page)(struct address_space *, sector_t,
542 /* migrate the contents of a page to the specified target */
543 int (*migratepage) (struct page *, struct page *);
544 int (*launder_page) (struct page *);
545 int (*error_remove_page) (struct mapping *mapping, struct page *page);
548 writepage: called by the VM to write a dirty page to backing store.
549 This may happen for data integrity reasons (i.e. 'sync'), or
550 to free up memory (flush). The difference can be seen in
552 The PG_Dirty flag has been cleared and PageLocked is true.
553 writepage should start writeout, should set PG_Writeback,
554 and should make sure the page is unlocked, either synchronously
555 or asynchronously when the write operation completes.
557 If wbc->sync_mode is WB_SYNC_NONE, ->writepage doesn't have to
558 try too hard if there are problems, and may choose to write out
559 other pages from the mapping if that is easier (e.g. due to
560 internal dependencies). If it chooses not to start writeout, it
561 should return AOP_WRITEPAGE_ACTIVATE so that the VM will not keep
562 calling ->writepage on that page.
564 See the file "Locking" for more details.
566 readpage: called by the VM to read a page from backing store.
567 The page will be Locked when readpage is called, and should be
568 unlocked and marked uptodate once the read completes.
569 If ->readpage discovers that it needs to unlock the page for
570 some reason, it can do so, and then return AOP_TRUNCATED_PAGE.
571 In this case, the page will be relocated, relocked and if
572 that all succeeds, ->readpage will be called again.
574 sync_page: called by the VM to notify the backing store to perform all
575 queued I/O operations for a page. I/O operations for other pages
576 associated with this address_space object may also be performed.
578 This function is optional and is called only for pages with
579 PG_Writeback set while waiting for the writeback to complete.
581 writepages: called by the VM to write out pages associated with the
582 address_space object. If wbc->sync_mode is WBC_SYNC_ALL, then
583 the writeback_control will specify a range of pages that must be
584 written out. If it is WBC_SYNC_NONE, then a nr_to_write is given
585 and that many pages should be written if possible.
586 If no ->writepages is given, then mpage_writepages is used
587 instead. This will choose pages from the address space that are
588 tagged as DIRTY and will pass them to ->writepage.
590 set_page_dirty: called by the VM to set a page dirty.
591 This is particularly needed if an address space attaches
592 private data to a page, and that data needs to be updated when
593 a page is dirtied. This is called, for example, when a memory
594 mapped page gets modified.
595 If defined, it should set the PageDirty flag, and the
596 PAGECACHE_TAG_DIRTY tag in the radix tree.
598 readpages: called by the VM to read pages associated with the address_space
599 object. This is essentially just a vector version of
600 readpage. Instead of just one page, several pages are
602 readpages is only used for read-ahead, so read errors are
603 ignored. If anything goes wrong, feel free to give up.
606 Called by the generic buffered write code to ask the filesystem to
607 prepare to write len bytes at the given offset in the file. The
608 address_space should check that the write will be able to complete,
609 by allocating space if necessary and doing any other internal
610 housekeeping. If the write will update parts of any basic-blocks on
611 storage, then those blocks should be pre-read (if they haven't been
612 read already) so that the updated blocks can be written out properly.
614 The filesystem must return the locked pagecache page for the specified
615 offset, in *pagep, for the caller to write into.
617 It must be able to cope with short writes (where the length passed to
618 write_begin is greater than the number of bytes copied into the page).
620 flags is a field for AOP_FLAG_xxx flags, described in
623 A void * may be returned in fsdata, which then gets passed into
626 Returns 0 on success; < 0 on failure (which is the error code), in
627 which case write_end is not called.
629 write_end: After a successful write_begin, and data copy, write_end must
630 be called. len is the original len passed to write_begin, and copied
631 is the amount that was able to be copied (copied == len is always true
632 if write_begin was called with the AOP_FLAG_UNINTERRUPTIBLE flag).
634 The filesystem must take care of unlocking the page and releasing it
635 refcount, and updating i_size.
637 Returns < 0 on failure, otherwise the number of bytes (<= 'copied')
638 that were able to be copied into pagecache.
640 bmap: called by the VFS to map a logical block offset within object to
641 physical block number. This method is used by the FIBMAP
642 ioctl and for working with swap-files. To be able to swap to
643 a file, the file must have a stable mapping to a block
644 device. The swap system does not go through the filesystem
645 but instead uses bmap to find out where the blocks in the file
646 are and uses those addresses directly.
649 invalidatepage: If a page has PagePrivate set, then invalidatepage
650 will be called when part or all of the page is to be removed
651 from the address space. This generally corresponds to either a
652 truncation or a complete invalidation of the address space
653 (in the latter case 'offset' will always be 0).
654 Any private data associated with the page should be updated
655 to reflect this truncation. If offset is 0, then
656 the private data should be released, because the page
657 must be able to be completely discarded. This may be done by
658 calling the ->releasepage function, but in this case the
659 release MUST succeed.
661 releasepage: releasepage is called on PagePrivate pages to indicate
662 that the page should be freed if possible. ->releasepage
663 should remove any private data from the page and clear the
664 PagePrivate flag. If releasepage() fails for some reason, it must
665 indicate failure with a 0 return value.
666 releasepage() is used in two distinct though related cases. The
667 first is when the VM finds a clean page with no active users and
668 wants to make it a free page. If ->releasepage succeeds, the
669 page will be removed from the address_space and become free.
671 The second case is when a request has been made to invalidate
672 some or all pages in an address_space. This can happen
673 through the fadvice(POSIX_FADV_DONTNEED) system call or by the
674 filesystem explicitly requesting it as nfs and 9fs do (when
675 they believe the cache may be out of date with storage) by
676 calling invalidate_inode_pages2().
677 If the filesystem makes such a call, and needs to be certain
678 that all pages are invalidated, then its releasepage will
679 need to ensure this. Possibly it can clear the PageUptodate
680 bit if it cannot free private data yet.
682 freepage: freepage is called once the page is no longer visible in
683 the page cache in order to allow the cleanup of any private
684 data. Since it may be called by the memory reclaimer, it
685 should not assume that the original address_space mapping still
686 exists, and it should not block.
688 direct_IO: called by the generic read/write routines to perform
689 direct_IO - that is IO requests which bypass the page cache
690 and transfer data directly between the storage and the
691 application's address space.
693 get_xip_page: called by the VM to translate a block number to a page.
694 The page is valid until the corresponding filesystem is unmounted.
695 Filesystems that want to use execute-in-place (XIP) need to implement
696 it. An example implementation can be found in fs/ext2/xip.c.
698 migrate_page: This is used to compact the physical memory usage.
699 If the VM wants to relocate a page (maybe off a memory card
700 that is signalling imminent failure) it will pass a new page
701 and an old page to this function. migrate_page should
702 transfer any private data across and update any references
703 that it has to the page.
705 launder_page: Called before freeing a page - it writes back the dirty page. To
706 prevent redirtying the page, it is kept locked during the whole
709 error_remove_page: normally set to generic_error_remove_page if truncation
710 is ok for this address space. Used for memory failure handling.
711 Setting this implies you deal with pages going away under you,
712 unless you have them locked or reference counts increased.
718 A file object represents a file opened by a process.
721 struct file_operations
722 ----------------------
724 This describes how the VFS can manipulate an open file. As of kernel
725 2.6.22, the following members are defined:
727 struct file_operations {
728 struct module *owner;
729 loff_t (*llseek) (struct file *, loff_t, int);
730 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
731 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
732 ssize_t (*aio_read) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
733 ssize_t (*aio_write) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
734 int (*readdir) (struct file *, void *, filldir_t);
735 unsigned int (*poll) (struct file *, struct poll_table_struct *);
736 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
737 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
738 int (*mmap) (struct file *, struct vm_area_struct *);
739 int (*open) (struct inode *, struct file *);
740 int (*flush) (struct file *);
741 int (*release) (struct inode *, struct file *);
742 int (*fsync) (struct file *, int datasync);
743 int (*aio_fsync) (struct kiocb *, int datasync);
744 int (*fasync) (int, struct file *, int);
745 int (*lock) (struct file *, int, struct file_lock *);
746 ssize_t (*readv) (struct file *, const struct iovec *, unsigned long, loff_t *);
747 ssize_t (*writev) (struct file *, const struct iovec *, unsigned long, loff_t *);
748 ssize_t (*sendfile) (struct file *, loff_t *, size_t, read_actor_t, void *);
749 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
750 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
751 int (*check_flags)(int);
752 int (*flock) (struct file *, int, struct file_lock *);
753 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, size_t, unsigned int);
754 ssize_t (*splice_read)(struct file *, struct pipe_inode_info *, size_t, unsigned int);
757 Again, all methods are called without any locks being held, unless
760 llseek: called when the VFS needs to move the file position index
762 read: called by read(2) and related system calls
764 aio_read: called by io_submit(2) and other asynchronous I/O operations
766 write: called by write(2) and related system calls
768 aio_write: called by io_submit(2) and other asynchronous I/O operations
770 readdir: called when the VFS needs to read the directory contents
772 poll: called by the VFS when a process wants to check if there is
773 activity on this file and (optionally) go to sleep until there
774 is activity. Called by the select(2) and poll(2) system calls
776 unlocked_ioctl: called by the ioctl(2) system call.
778 compat_ioctl: called by the ioctl(2) system call when 32 bit system calls
779 are used on 64 bit kernels.
781 mmap: called by the mmap(2) system call
783 open: called by the VFS when an inode should be opened. When the VFS
784 opens a file, it creates a new "struct file". It then calls the
785 open method for the newly allocated file structure. You might
786 think that the open method really belongs in
787 "struct inode_operations", and you may be right. I think it's
788 done the way it is because it makes filesystems simpler to
789 implement. The open() method is a good place to initialize the
790 "private_data" member in the file structure if you want to point
791 to a device structure
793 flush: called by the close(2) system call to flush a file
795 release: called when the last reference to an open file is closed
797 fsync: called by the fsync(2) system call
799 fasync: called by the fcntl(2) system call when asynchronous
800 (non-blocking) mode is enabled for a file
802 lock: called by the fcntl(2) system call for F_GETLK, F_SETLK, and F_SETLKW
805 readv: called by the readv(2) system call
807 writev: called by the writev(2) system call
809 sendfile: called by the sendfile(2) system call
811 get_unmapped_area: called by the mmap(2) system call
813 check_flags: called by the fcntl(2) system call for F_SETFL command
815 flock: called by the flock(2) system call
817 splice_write: called by the VFS to splice data from a pipe to a file. This
818 method is used by the splice(2) system call
820 splice_read: called by the VFS to splice data from file to a pipe. This
821 method is used by the splice(2) system call
823 Note that the file operations are implemented by the specific
824 filesystem in which the inode resides. When opening a device node
825 (character or block special) most filesystems will call special
826 support routines in the VFS which will locate the required device
827 driver information. These support routines replace the filesystem file
828 operations with those for the device driver, and then proceed to call
829 the new open() method for the file. This is how opening a device file
830 in the filesystem eventually ends up calling the device driver open()
834 Directory Entry Cache (dcache)
835 ==============================
838 struct dentry_operations
839 ------------------------
841 This describes how a filesystem can overload the standard dentry
842 operations. Dentries and the dcache are the domain of the VFS and the
843 individual filesystem implementations. Device drivers have no business
844 here. These methods may be set to NULL, as they are either optional or
845 the VFS uses a default. As of kernel 2.6.22, the following members are
848 struct dentry_operations {
849 int (*d_revalidate)(struct dentry *, struct nameidata *);
850 int (*d_hash) (struct dentry *, struct qstr *);
851 int (*d_compare) (struct dentry *, struct qstr *, struct qstr *);
852 int (*d_delete)(struct dentry *);
853 void (*d_release)(struct dentry *);
854 void (*d_iput)(struct dentry *, struct inode *);
855 char *(*d_dname)(struct dentry *, char *, int);
858 d_revalidate: called when the VFS needs to revalidate a dentry. This
859 is called whenever a name look-up finds a dentry in the
860 dcache. Most filesystems leave this as NULL, because all their
861 dentries in the dcache are valid
863 d_hash: called when the VFS adds a dentry to the hash table
865 d_compare: called when a dentry should be compared with another
867 d_delete: called when the last reference to a dentry is
868 deleted. This means no-one is using the dentry, however it is
869 still valid and in the dcache
871 d_release: called when a dentry is really deallocated
873 d_iput: called when a dentry loses its inode (just prior to its
874 being deallocated). The default when this is NULL is that the
875 VFS calls iput(). If you define this method, you must call
878 d_dname: called when the pathname of a dentry should be generated.
879 Useful for some pseudo filesystems (sockfs, pipefs, ...) to delay
880 pathname generation. (Instead of doing it when dentry is created,
881 it's done only when the path is needed.). Real filesystems probably
882 dont want to use it, because their dentries are present in global
883 dcache hash, so their hash should be an invariant. As no lock is
884 held, d_dname() should not try to modify the dentry itself, unless
885 appropriate SMP safety is used. CAUTION : d_path() logic is quite
886 tricky. The correct way to return for example "Hello" is to put it
887 at the end of the buffer, and returns a pointer to the first char.
888 dynamic_dname() helper function is provided to take care of this.
892 static char *pipefs_dname(struct dentry *dent, char *buffer, int buflen)
894 return dynamic_dname(dentry, buffer, buflen, "pipe:[%lu]",
895 dentry->d_inode->i_ino);
898 Each dentry has a pointer to its parent dentry, as well as a hash list
899 of child dentries. Child dentries are basically like files in a
903 Directory Entry Cache API
904 --------------------------
906 There are a number of functions defined which permit a filesystem to
909 dget: open a new handle for an existing dentry (this just increments
912 dput: close a handle for a dentry (decrements the usage count). If
913 the usage count drops to 0, the "d_delete" method is called
914 and the dentry is placed on the unused list if the dentry is
915 still in its parents hash list. Putting the dentry on the
916 unused list just means that if the system needs some RAM, it
917 goes through the unused list of dentries and deallocates them.
918 If the dentry has already been unhashed and the usage count
919 drops to 0, in this case the dentry is deallocated after the
920 "d_delete" method is called
922 d_drop: this unhashes a dentry from its parents hash list. A
923 subsequent call to dput() will deallocate the dentry if its
924 usage count drops to 0
926 d_delete: delete a dentry. If there are no other open references to
927 the dentry then the dentry is turned into a negative dentry
928 (the d_iput() method is called). If there are other
929 references, then d_drop() is called instead
931 d_add: add a dentry to its parents hash list and then calls
934 d_instantiate: add a dentry to the alias hash list for the inode and
935 updates the "d_inode" member. The "i_count" member in the
936 inode structure should be set/incremented. If the inode
937 pointer is NULL, the dentry is called a "negative
938 dentry". This function is commonly called when an inode is
939 created for an existing negative dentry
941 d_lookup: look up a dentry given its parent and path name component
942 It looks up the child of that given name from the dcache
943 hash table. If it is found, the reference count is incremented
944 and the dentry is returned. The caller must use dput()
945 to free the dentry when it finishes using it.
947 For further information on dentry locking, please refer to the document
948 Documentation/filesystems/dentry-locking.txt.
956 On mount and remount the filesystem is passed a string containing a
957 comma separated list of mount options. The options can have either of
963 The <linux/parser.h> header defines an API that helps parse these
964 options. There are plenty of examples on how to use it in existing
970 If a filesystem accepts mount options, it must define show_options()
971 to show all the currently active options. The rules are:
973 - options MUST be shown which are not default or their values differ
976 - options MAY be shown which are enabled by default or have their
979 Options used only internally between a mount helper and the kernel
980 (such as file descriptors), or which only have an effect during the
981 mounting (such as ones controlling the creation of a journal) are exempt
982 from the above rules.
984 The underlying reason for the above rules is to make sure, that a
985 mount can be accurately replicated (e.g. umounting and mounting again)
986 based on the information found in /proc/mounts.
988 A simple method of saving options at mount/remount time and showing
989 them is provided with the save_mount_options() and
990 generic_show_options() helper functions. Please note, that using
991 these may have drawbacks. For more info see header comments for these
992 functions in fs/namespace.c.
997 (Note some of these resources are not up-to-date with the latest kernel
1000 Creating Linux virtual filesystems. 2002
1001 <http://lwn.net/Articles/13325/>
1003 The Linux Virtual File-system Layer by Neil Brown. 1999
1004 <http://www.cse.unsw.edu.au/~neilb/oss/linux-commentary/vfs.html>
1006 A tour of the Linux VFS by Michael K. Johnson. 1996
1007 <http://www.tldp.org/LDP/khg/HyperNews/get/fs/vfstour.html>
1009 A small trail through the Linux kernel by Andries Brouwer. 2001
1010 <http://www.win.tue.nl/~aeb/linux/vfs/trail.html>