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 it's 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: decribes 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
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
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 (*put_inode) (struct inode *);
209 void (*drop_inode) (struct inode *);
210 void (*delete_inode) (struct inode *);
211 void (*put_super) (struct super_block *);
212 void (*write_super) (struct super_block *);
213 int (*sync_fs)(struct super_block *sb, int wait);
214 void (*write_super_lockfs) (struct super_block *);
215 void (*unlockfs) (struct super_block *);
216 int (*statfs) (struct dentry *, struct kstatfs *);
217 int (*remount_fs) (struct super_block *, int *, char *);
218 void (*clear_inode) (struct inode *);
219 void (*umount_begin) (struct super_block *);
221 int (*show_options)(struct seq_file *, struct vfsmount *);
223 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
224 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
227 All methods are called without any locks being held, unless otherwise
228 noted. This means that most methods can block safely. All methods are
229 only called from a process context (i.e. not from an interrupt handler
232 alloc_inode: this method is called by inode_alloc() to allocate memory
233 for struct inode and initialize it. If this function is not
234 defined, a simple 'struct inode' is allocated. Normally
235 alloc_inode will be used to allocate a larger structure which
236 contains a 'struct inode' embedded within it.
238 destroy_inode: this method is called by destroy_inode() to release
239 resources allocated for struct inode. It is only required if
240 ->alloc_inode was defined and simply undoes anything done by
243 dirty_inode: this method is called by the VFS to mark an inode dirty.
245 write_inode: this method is called when the VFS needs to write an
246 inode to disc. The second parameter indicates whether the write
247 should be synchronous or not, not all filesystems check this flag.
249 put_inode: called when the VFS inode is removed from the inode
252 drop_inode: called when the last access to the inode is dropped,
253 with the inode_lock spinlock held.
255 This method should be either NULL (normal UNIX filesystem
256 semantics) or "generic_delete_inode" (for filesystems that do not
257 want to cache inodes - causing "delete_inode" to always be
258 called regardless of the value of i_nlink)
260 The "generic_delete_inode()" behavior is equivalent to the
261 old practice of using "force_delete" in the put_inode() case,
262 but does not have the races that the "force_delete()" approach
265 delete_inode: called when the VFS wants to delete an inode
267 put_super: called when the VFS wishes to free the superblock
268 (i.e. unmount). This is called with the superblock lock held
270 write_super: called when the VFS superblock needs to be written to
271 disc. This method is optional
273 sync_fs: called when VFS is writing out all dirty data associated with
274 a superblock. The second parameter indicates whether the method
275 should wait until the write out has been completed. Optional.
277 write_super_lockfs: called when VFS is locking a filesystem and
278 forcing it into a consistent state. This method is currently
279 used by the Logical Volume Manager (LVM).
281 unlockfs: called when VFS is unlocking a filesystem and making it writable
284 statfs: called when the VFS needs to get filesystem statistics. This
285 is called with the kernel lock held
287 remount_fs: called when the filesystem is remounted. This is called
288 with the kernel lock held
290 clear_inode: called then the VFS clears the inode. Optional
292 umount_begin: called when the VFS is unmounting a filesystem.
294 show_options: called by the VFS to show mount options for /proc/<pid>/mounts.
296 quota_read: called by the VFS to read from filesystem quota file.
298 quota_write: called by the VFS to write to filesystem quota file.
300 Whoever sets up the inode is responsible for filling in the "i_op" field. This
301 is a pointer to a "struct inode_operations" which describes the methods that
302 can be performed on individual inodes.
308 An inode object represents an object within the filesystem.
311 struct inode_operations
312 -----------------------
314 This describes how the VFS can manipulate an inode in your
315 filesystem. As of kernel 2.6.22, the following members are defined:
317 struct inode_operations {
318 int (*create) (struct inode *,struct dentry *,int, struct nameidata *);
319 struct dentry * (*lookup) (struct inode *,struct dentry *, struct nameidata *);
320 int (*link) (struct dentry *,struct inode *,struct dentry *);
321 int (*unlink) (struct inode *,struct dentry *);
322 int (*symlink) (struct inode *,struct dentry *,const char *);
323 int (*mkdir) (struct inode *,struct dentry *,int);
324 int (*rmdir) (struct inode *,struct dentry *);
325 int (*mknod) (struct inode *,struct dentry *,int,dev_t);
326 int (*rename) (struct inode *, struct dentry *,
327 struct inode *, struct dentry *);
328 int (*readlink) (struct dentry *, char __user *,int);
329 void * (*follow_link) (struct dentry *, struct nameidata *);
330 void (*put_link) (struct dentry *, struct nameidata *, void *);
331 void (*truncate) (struct inode *);
332 int (*permission) (struct inode *, int, struct nameidata *);
333 int (*setattr) (struct dentry *, struct iattr *);
334 int (*getattr) (struct vfsmount *mnt, struct dentry *, struct kstat *);
335 int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
336 ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t);
337 ssize_t (*listxattr) (struct dentry *, char *, size_t);
338 int (*removexattr) (struct dentry *, const char *);
339 void (*truncate_range)(struct inode *, loff_t, loff_t);
342 Again, all methods are called without any locks being held, unless
345 create: called by the open(2) and creat(2) system calls. Only
346 required if you want to support regular files. The dentry you
347 get should not have an inode (i.e. it should be a negative
348 dentry). Here you will probably call d_instantiate() with the
349 dentry and the newly created inode
351 lookup: called when the VFS needs to look up an inode in a parent
352 directory. The name to look for is found in the dentry. This
353 method must call d_add() to insert the found inode into the
354 dentry. The "i_count" field in the inode structure should be
355 incremented. If the named inode does not exist a NULL inode
356 should be inserted into the dentry (this is called a negative
357 dentry). Returning an error code from this routine must only
358 be done on a real error, otherwise creating inodes with system
359 calls like create(2), mknod(2), mkdir(2) and so on will fail.
360 If you wish to overload the dentry methods then you should
361 initialise the "d_dop" field in the dentry; this is a pointer
362 to a struct "dentry_operations".
363 This method is called with the directory inode semaphore held
365 link: called by the link(2) system call. Only required if you want
366 to support hard links. You will probably need to call
367 d_instantiate() just as you would in the create() method
369 unlink: called by the unlink(2) system call. Only required if you
370 want to support deleting inodes
372 symlink: called by the symlink(2) system call. Only required if you
373 want to support symlinks. You will probably need to call
374 d_instantiate() just as you would in the create() method
376 mkdir: called by the mkdir(2) system call. Only required if you want
377 to support creating subdirectories. You will probably need to
378 call d_instantiate() just as you would in the create() method
380 rmdir: called by the rmdir(2) system call. Only required if you want
381 to support deleting subdirectories
383 mknod: called by the mknod(2) system call to create a device (char,
384 block) inode or a named pipe (FIFO) or socket. Only required
385 if you want to support creating these types of inodes. You
386 will probably need to call d_instantiate() just as you would
387 in the create() method
389 rename: called by the rename(2) system call to rename the object to
390 have the parent and name given by the second inode and dentry.
392 readlink: called by the readlink(2) system call. Only required if
393 you want to support reading symbolic links
395 follow_link: called by the VFS to follow a symbolic link to the
396 inode it points to. Only required if you want to support
397 symbolic links. This method returns a void pointer cookie
398 that is passed to put_link().
400 put_link: called by the VFS to release resources allocated by
401 follow_link(). The cookie returned by follow_link() is passed
402 to this method as the last parameter. It is used by
403 filesystems such as NFS where page cache is not stable
404 (i.e. page that was installed when the symbolic link walk
405 started might not be in the page cache at the end of the
408 truncate: called by the VFS to change the size of a file. The
409 i_size field of the inode is set to the desired size by the
410 VFS before this method is called. This method is called by
411 the truncate(2) system call and related functionality.
413 permission: called by the VFS to check for access rights on a POSIX-like
416 setattr: called by the VFS to set attributes for a file. This method
417 is called by chmod(2) and related system calls.
419 getattr: called by the VFS to get attributes of a file. This method
420 is called by stat(2) and related system calls.
422 setxattr: called by the VFS to set an extended attribute for a file.
423 Extended attribute is a name:value pair associated with an
424 inode. This method is called by setxattr(2) system call.
426 getxattr: called by the VFS to retrieve the value of an extended
427 attribute name. This method is called by getxattr(2) function
430 listxattr: called by the VFS to list all extended attributes for a
431 given file. This method is called by listxattr(2) system call.
433 removexattr: called by the VFS to remove an extended attribute from
434 a file. This method is called by removexattr(2) system call.
436 truncate_range: a method provided by the underlying filesystem to truncate a
437 range of blocks , i.e. punch a hole somewhere in a file.
440 The Address Space Object
441 ========================
443 The address space object is used to group and manage pages in the page
444 cache. It can be used to keep track of the pages in a file (or
445 anything else) and also track the mapping of sections of the file into
446 process address spaces.
448 There are a number of distinct yet related services that an
449 address-space can provide. These include communicating memory
450 pressure, page lookup by address, and keeping track of pages tagged as
453 The first can be used independently to the others. The VM can try to
454 either write dirty pages in order to clean them, or release clean
455 pages in order to reuse them. To do this it can call the ->writepage
456 method on dirty pages, and ->releasepage on clean pages with
457 PagePrivate set. Clean pages without PagePrivate and with no external
458 references will be released without notice being given to the
461 To achieve this functionality, pages need to be placed on an LRU with
462 lru_cache_add and mark_page_active needs to be called whenever the
465 Pages are normally kept in a radix tree index by ->index. This tree
466 maintains information about the PG_Dirty and PG_Writeback status of
467 each page, so that pages with either of these flags can be found
470 The Dirty tag is primarily used by mpage_writepages - the default
471 ->writepages method. It uses the tag to find dirty pages to call
472 ->writepage on. If mpage_writepages is not used (i.e. the address
473 provides its own ->writepages) , the PAGECACHE_TAG_DIRTY tag is
474 almost unused. write_inode_now and sync_inode do use it (through
475 __sync_single_inode) to check if ->writepages has been successful in
476 writing out the whole address_space.
478 The Writeback tag is used by filemap*wait* and sync_page* functions,
479 via wait_on_page_writeback_range, to wait for all writeback to
480 complete. While waiting ->sync_page (if defined) will be called on
481 each page that is found to require writeback.
483 An address_space handler may attach extra information to a page,
484 typically using the 'private' field in the 'struct page'. If such
485 information is attached, the PG_Private flag should be set. This will
486 cause various VM routines to make extra calls into the address_space
487 handler to deal with that data.
489 An address space acts as an intermediate between storage and
490 application. Data is read into the address space a whole page at a
491 time, and provided to the application either by copying of the page,
492 or by memory-mapping the page.
493 Data is written into the address space by the application, and then
494 written-back to storage typically in whole pages, however the
495 address_space has finer control of write sizes.
497 The read process essentially only requires 'readpage'. The write
498 process is more complicated and uses prepare_write/commit_write or
499 set_page_dirty to write data into the address_space, and writepage,
500 sync_page, and writepages to writeback data to storage.
502 Adding and removing pages to/from an address_space is protected by the
505 When data is written to a page, the PG_Dirty flag should be set. It
506 typically remains set until writepage asks for it to be written. This
507 should clear PG_Dirty and set PG_Writeback. It can be actually
508 written at any point after PG_Dirty is clear. Once it is known to be
509 safe, PG_Writeback is cleared.
511 Writeback makes use of a writeback_control structure...
513 struct address_space_operations
514 -------------------------------
516 This describes how the VFS can manipulate mapping of a file to page cache in
517 your filesystem. As of kernel 2.6.22, the following members are defined:
519 struct address_space_operations {
520 int (*writepage)(struct page *page, struct writeback_control *wbc);
521 int (*readpage)(struct file *, struct page *);
522 int (*sync_page)(struct page *);
523 int (*writepages)(struct address_space *, struct writeback_control *);
524 int (*set_page_dirty)(struct page *page);
525 int (*readpages)(struct file *filp, struct address_space *mapping,
526 struct list_head *pages, unsigned nr_pages);
527 int (*prepare_write)(struct file *, struct page *, unsigned, unsigned);
528 int (*commit_write)(struct file *, struct page *, unsigned, unsigned);
529 int (*write_begin)(struct file *, struct address_space *mapping,
530 loff_t pos, unsigned len, unsigned flags,
531 struct page **pagep, void **fsdata);
532 int (*write_end)(struct file *, struct address_space *mapping,
533 loff_t pos, unsigned len, unsigned copied,
534 struct page *page, void *fsdata);
535 sector_t (*bmap)(struct address_space *, sector_t);
536 int (*invalidatepage) (struct page *, unsigned long);
537 int (*releasepage) (struct page *, int);
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 *);
547 writepage: called by the VM to write a dirty page to backing store.
548 This may happen for data integrity reasons (i.e. 'sync'), or
549 to free up memory (flush). The difference can be seen in
551 The PG_Dirty flag has been cleared and PageLocked is true.
552 writepage should start writeout, should set PG_Writeback,
553 and should make sure the page is unlocked, either synchronously
554 or asynchronously when the write operation completes.
556 If wbc->sync_mode is WB_SYNC_NONE, ->writepage doesn't have to
557 try too hard if there are problems, and may choose to write out
558 other pages from the mapping if that is easier (e.g. due to
559 internal dependencies). If it chooses not to start writeout, it
560 should return AOP_WRITEPAGE_ACTIVATE so that the VM will not keep
561 calling ->writepage on that page.
563 See the file "Locking" for more details.
565 readpage: called by the VM to read a page from backing store.
566 The page will be Locked when readpage is called, and should be
567 unlocked and marked uptodate once the read completes.
568 If ->readpage discovers that it needs to unlock the page for
569 some reason, it can do so, and then return AOP_TRUNCATED_PAGE.
570 In this case, the page will be relocated, relocked and if
571 that all succeeds, ->readpage will be called again.
573 sync_page: called by the VM to notify the backing store to perform all
574 queued I/O operations for a page. I/O operations for other pages
575 associated with this address_space object may also be performed.
577 This function is optional and is called only for pages with
578 PG_Writeback set while waiting for the writeback to complete.
580 writepages: called by the VM to write out pages associated with the
581 address_space object. If wbc->sync_mode is WBC_SYNC_ALL, then
582 the writeback_control will specify a range of pages that must be
583 written out. If it is WBC_SYNC_NONE, then a nr_to_write is given
584 and that many pages should be written if possible.
585 If no ->writepages is given, then mpage_writepages is used
586 instead. This will choose pages from the address space that are
587 tagged as DIRTY and will pass them to ->writepage.
589 set_page_dirty: called by the VM to set a page dirty.
590 This is particularly needed if an address space attaches
591 private data to a page, and that data needs to be updated when
592 a page is dirtied. This is called, for example, when a memory
593 mapped page gets modified.
594 If defined, it should set the PageDirty flag, and the
595 PAGECACHE_TAG_DIRTY tag in the radix tree.
597 readpages: called by the VM to read pages associated with the address_space
598 object. This is essentially just a vector version of
599 readpage. Instead of just one page, several pages are
601 readpages is only used for read-ahead, so read errors are
602 ignored. If anything goes wrong, feel free to give up.
604 prepare_write: called by the generic write path in VM to set up a write
605 request for a page. This indicates to the address space that
606 the given range of bytes is about to be written. The
607 address_space should check that the write will be able to
608 complete, by allocating space if necessary and doing any other
609 internal housekeeping. If the write will update parts of
610 any basic-blocks on storage, then those blocks should be
611 pre-read (if they haven't been read already) so that the
612 updated blocks can be written out properly.
613 The page will be locked.
615 Note: the page _must not_ be marked uptodate in this function
616 (or anywhere else) unless it actually is uptodate right now. As
617 soon as a page is marked uptodate, it is possible for a concurrent
618 read(2) to copy it to userspace.
620 commit_write: If prepare_write succeeds, new data will be copied
621 into the page and then commit_write will be called. It will
622 typically update the size of the file (if appropriate) and
623 mark the inode as dirty, and do any other related housekeeping
624 operations. It should avoid returning an error if possible -
625 errors should have been handled by prepare_write.
627 write_begin: This is intended as a replacement for prepare_write. The
628 key differences being that:
629 - it returns a locked page (in *pagep) rather than being
630 given a pre locked page;
631 - it must be able to cope with short writes (where the
632 length passed to write_begin is greater than the number
633 of bytes copied into the page).
635 Called by the generic buffered write code to ask the filesystem to
636 prepare to write len bytes at the given offset in the file. The
637 address_space should check that the write will be able to complete,
638 by allocating space if necessary and doing any other internal
639 housekeeping. If the write will update parts of any basic-blocks on
640 storage, then those blocks should be pre-read (if they haven't been
641 read already) so that the updated blocks can be written out properly.
643 The filesystem must return the locked pagecache page for the specified
644 offset, in *pagep, for the caller to write into.
646 flags is a field for AOP_FLAG_xxx flags, described in
649 A void * may be returned in fsdata, which then gets passed into
652 Returns 0 on success; < 0 on failure (which is the error code), in
653 which case write_end is not called.
655 write_end: After a successful write_begin, and data copy, write_end must
656 be called. len is the original len passed to write_begin, and copied
657 is the amount that was able to be copied (copied == len is always true
658 if write_begin was called with the AOP_FLAG_UNINTERRUPTIBLE flag).
660 The filesystem must take care of unlocking the page and releasing it
661 refcount, and updating i_size.
663 Returns < 0 on failure, otherwise the number of bytes (<= 'copied')
664 that were able to be copied into pagecache.
666 bmap: called by the VFS to map a logical block offset within object to
667 physical block number. This method is used by the FIBMAP
668 ioctl and for working with swap-files. To be able to swap to
669 a file, the file must have a stable mapping to a block
670 device. The swap system does not go through the filesystem
671 but instead uses bmap to find out where the blocks in the file
672 are and uses those addresses directly.
675 invalidatepage: If a page has PagePrivate set, then invalidatepage
676 will be called when part or all of the page is to be removed
677 from the address space. This generally corresponds to either a
678 truncation or a complete invalidation of the address space
679 (in the latter case 'offset' will always be 0).
680 Any private data associated with the page should be updated
681 to reflect this truncation. If offset is 0, then
682 the private data should be released, because the page
683 must be able to be completely discarded. This may be done by
684 calling the ->releasepage function, but in this case the
685 release MUST succeed.
687 releasepage: releasepage is called on PagePrivate pages to indicate
688 that the page should be freed if possible. ->releasepage
689 should remove any private data from the page and clear the
690 PagePrivate flag. It may also remove the page from the
691 address_space. If this fails for some reason, it may indicate
692 failure with a 0 return value.
693 This is used in two distinct though related cases. The first
694 is when the VM finds a clean page with no active users and
695 wants to make it a free page. If ->releasepage succeeds, the
696 page will be removed from the address_space and become free.
698 The second case is when a request has been made to invalidate
699 some or all pages in an address_space. This can happen
700 through the fadvice(POSIX_FADV_DONTNEED) system call or by the
701 filesystem explicitly requesting it as nfs and 9fs do (when
702 they believe the cache may be out of date with storage) by
703 calling invalidate_inode_pages2().
704 If the filesystem makes such a call, and needs to be certain
705 that all pages are invalidated, then its releasepage will
706 need to ensure this. Possibly it can clear the PageUptodate
707 bit if it cannot free private data yet.
709 direct_IO: called by the generic read/write routines to perform
710 direct_IO - that is IO requests which bypass the page cache
711 and transfer data directly between the storage and the
712 application's address space.
714 get_xip_page: called by the VM to translate a block number to a page.
715 The page is valid until the corresponding filesystem is unmounted.
716 Filesystems that want to use execute-in-place (XIP) need to implement
717 it. An example implementation can be found in fs/ext2/xip.c.
719 migrate_page: This is used to compact the physical memory usage.
720 If the VM wants to relocate a page (maybe off a memory card
721 that is signalling imminent failure) it will pass a new page
722 and an old page to this function. migrate_page should
723 transfer any private data across and update any references
724 that it has to the page.
726 launder_page: Called before freeing a page - it writes back the dirty page. To
727 prevent redirtying the page, it is kept locked during the whole
733 A file object represents a file opened by a process.
736 struct file_operations
737 ----------------------
739 This describes how the VFS can manipulate an open file. As of kernel
740 2.6.22, the following members are defined:
742 struct file_operations {
743 struct module *owner;
744 loff_t (*llseek) (struct file *, loff_t, int);
745 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
746 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
747 ssize_t (*aio_read) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
748 ssize_t (*aio_write) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
749 int (*readdir) (struct file *, void *, filldir_t);
750 unsigned int (*poll) (struct file *, struct poll_table_struct *);
751 int (*ioctl) (struct inode *, struct file *, unsigned int, unsigned long);
752 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
753 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
754 int (*mmap) (struct file *, struct vm_area_struct *);
755 int (*open) (struct inode *, struct file *);
756 int (*flush) (struct file *);
757 int (*release) (struct inode *, struct file *);
758 int (*fsync) (struct file *, struct dentry *, int datasync);
759 int (*aio_fsync) (struct kiocb *, int datasync);
760 int (*fasync) (int, struct file *, int);
761 int (*lock) (struct file *, int, struct file_lock *);
762 ssize_t (*readv) (struct file *, const struct iovec *, unsigned long, loff_t *);
763 ssize_t (*writev) (struct file *, const struct iovec *, unsigned long, loff_t *);
764 ssize_t (*sendfile) (struct file *, loff_t *, size_t, read_actor_t, void *);
765 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
766 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
767 int (*check_flags)(int);
768 int (*dir_notify)(struct file *filp, unsigned long arg);
769 int (*flock) (struct file *, int, struct file_lock *);
770 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, size_t, unsigned int);
771 ssize_t (*splice_read)(struct file *, struct pipe_inode_info *, size_t, unsigned int);
774 Again, all methods are called without any locks being held, unless
777 llseek: called when the VFS needs to move the file position index
779 read: called by read(2) and related system calls
781 aio_read: called by io_submit(2) and other asynchronous I/O operations
783 write: called by write(2) and related system calls
785 aio_write: called by io_submit(2) and other asynchronous I/O operations
787 readdir: called when the VFS needs to read the directory contents
789 poll: called by the VFS when a process wants to check if there is
790 activity on this file and (optionally) go to sleep until there
791 is activity. Called by the select(2) and poll(2) system calls
793 ioctl: called by the ioctl(2) system call
795 unlocked_ioctl: called by the ioctl(2) system call. Filesystems that do not
796 require the BKL should use this method instead of the ioctl() above.
798 compat_ioctl: called by the ioctl(2) system call when 32 bit system calls
799 are used on 64 bit kernels.
801 mmap: called by the mmap(2) system call
803 open: called by the VFS when an inode should be opened. When the VFS
804 opens a file, it creates a new "struct file". It then calls the
805 open method for the newly allocated file structure. You might
806 think that the open method really belongs in
807 "struct inode_operations", and you may be right. I think it's
808 done the way it is because it makes filesystems simpler to
809 implement. The open() method is a good place to initialize the
810 "private_data" member in the file structure if you want to point
811 to a device structure
813 flush: called by the close(2) system call to flush a file
815 release: called when the last reference to an open file is closed
817 fsync: called by the fsync(2) system call
819 fasync: called by the fcntl(2) system call when asynchronous
820 (non-blocking) mode is enabled for a file
822 lock: called by the fcntl(2) system call for F_GETLK, F_SETLK, and F_SETLKW
825 readv: called by the readv(2) system call
827 writev: called by the writev(2) system call
829 sendfile: called by the sendfile(2) system call
831 get_unmapped_area: called by the mmap(2) system call
833 check_flags: called by the fcntl(2) system call for F_SETFL command
835 dir_notify: called by the fcntl(2) system call for F_NOTIFY command
837 flock: called by the flock(2) system call
839 splice_write: called by the VFS to splice data from a pipe to a file. This
840 method is used by the splice(2) system call
842 splice_read: called by the VFS to splice data from file to a pipe. This
843 method is used by the splice(2) system call
845 Note that the file operations are implemented by the specific
846 filesystem in which the inode resides. When opening a device node
847 (character or block special) most filesystems will call special
848 support routines in the VFS which will locate the required device
849 driver information. These support routines replace the filesystem file
850 operations with those for the device driver, and then proceed to call
851 the new open() method for the file. This is how opening a device file
852 in the filesystem eventually ends up calling the device driver open()
856 Directory Entry Cache (dcache)
857 ==============================
860 struct dentry_operations
861 ------------------------
863 This describes how a filesystem can overload the standard dentry
864 operations. Dentries and the dcache are the domain of the VFS and the
865 individual filesystem implementations. Device drivers have no business
866 here. These methods may be set to NULL, as they are either optional or
867 the VFS uses a default. As of kernel 2.6.22, the following members are
870 struct dentry_operations {
871 int (*d_revalidate)(struct dentry *, struct nameidata *);
872 int (*d_hash) (struct dentry *, struct qstr *);
873 int (*d_compare) (struct dentry *, struct qstr *, struct qstr *);
874 int (*d_delete)(struct dentry *);
875 void (*d_release)(struct dentry *);
876 void (*d_iput)(struct dentry *, struct inode *);
877 char *(*d_dname)(struct dentry *, char *, int);
880 d_revalidate: called when the VFS needs to revalidate a dentry. This
881 is called whenever a name look-up finds a dentry in the
882 dcache. Most filesystems leave this as NULL, because all their
883 dentries in the dcache are valid
885 d_hash: called when the VFS adds a dentry to the hash table
887 d_compare: called when a dentry should be compared with another
889 d_delete: called when the last reference to a dentry is
890 deleted. This means no-one is using the dentry, however it is
891 still valid and in the dcache
893 d_release: called when a dentry is really deallocated
895 d_iput: called when a dentry loses its inode (just prior to its
896 being deallocated). The default when this is NULL is that the
897 VFS calls iput(). If you define this method, you must call
900 d_dname: called when the pathname of a dentry should be generated.
901 Usefull for some pseudo filesystems (sockfs, pipefs, ...) to delay
902 pathname generation. (Instead of doing it when dentry is created,
903 its done only when the path is needed.). Real filesystems probably
904 dont want to use it, because their dentries are present in global
905 dcache hash, so their hash should be an invariant. As no lock is
906 held, d_dname() should not try to modify the dentry itself, unless
907 appropriate SMP safety is used. CAUTION : d_path() logic is quite
908 tricky. The correct way to return for example "Hello" is to put it
909 at the end of the buffer, and returns a pointer to the first char.
910 dynamic_dname() helper function is provided to take care of this.
914 static char *pipefs_dname(struct dentry *dent, char *buffer, int buflen)
916 return dynamic_dname(dentry, buffer, buflen, "pipe:[%lu]",
917 dentry->d_inode->i_ino);
920 Each dentry has a pointer to its parent dentry, as well as a hash list
921 of child dentries. Child dentries are basically like files in a
925 Directory Entry Cache API
926 --------------------------
928 There are a number of functions defined which permit a filesystem to
931 dget: open a new handle for an existing dentry (this just increments
934 dput: close a handle for a dentry (decrements the usage count). If
935 the usage count drops to 0, the "d_delete" method is called
936 and the dentry is placed on the unused list if the dentry is
937 still in its parents hash list. Putting the dentry on the
938 unused list just means that if the system needs some RAM, it
939 goes through the unused list of dentries and deallocates them.
940 If the dentry has already been unhashed and the usage count
941 drops to 0, in this case the dentry is deallocated after the
942 "d_delete" method is called
944 d_drop: this unhashes a dentry from its parents hash list. A
945 subsequent call to dput() will deallocate the dentry if its
946 usage count drops to 0
948 d_delete: delete a dentry. If there are no other open references to
949 the dentry then the dentry is turned into a negative dentry
950 (the d_iput() method is called). If there are other
951 references, then d_drop() is called instead
953 d_add: add a dentry to its parents hash list and then calls
956 d_instantiate: add a dentry to the alias hash list for the inode and
957 updates the "d_inode" member. The "i_count" member in the
958 inode structure should be set/incremented. If the inode
959 pointer is NULL, the dentry is called a "negative
960 dentry". This function is commonly called when an inode is
961 created for an existing negative dentry
963 d_lookup: look up a dentry given its parent and path name component
964 It looks up the child of that given name from the dcache
965 hash table. If it is found, the reference count is incremented
966 and the dentry is returned. The caller must use d_put()
967 to free the dentry when it finishes using it.
969 For further information on dentry locking, please refer to the document
970 Documentation/filesystems/dentry-locking.txt.
976 (Note some of these resources are not up-to-date with the latest kernel
979 Creating Linux virtual filesystems. 2002
980 <http://lwn.net/Articles/13325/>
982 The Linux Virtual File-system Layer by Neil Brown. 1999
983 <http://www.cse.unsw.edu.au/~neilb/oss/linux-commentary/vfs.html>
985 A tour of the Linux VFS by Michael K. Johnson. 1996
986 <http://www.tldp.org/LDP/khg/HyperNews/get/fs/vfstour.html>
988 A small trail through the Linux kernel by Andries Brouwer. 2001
989 <http://www.win.tue.nl/~aeb/linux/vfs/trail.html>