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 2.6.13, 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 (*read_inode) (struct inode *);
208 void (*dirty_inode) (struct inode *);
209 int (*write_inode) (struct inode *, int);
210 void (*put_inode) (struct inode *);
211 void (*drop_inode) (struct inode *);
212 void (*delete_inode) (struct inode *);
213 void (*put_super) (struct super_block *);
214 void (*write_super) (struct super_block *);
215 int (*sync_fs)(struct super_block *sb, int wait);
216 void (*write_super_lockfs) (struct super_block *);
217 void (*unlockfs) (struct super_block *);
218 int (*statfs) (struct dentry *, struct kstatfs *);
219 int (*remount_fs) (struct super_block *, int *, char *);
220 void (*clear_inode) (struct inode *);
221 void (*umount_begin) (struct super_block *);
223 void (*sync_inodes) (struct super_block *sb,
224 struct writeback_control *wbc);
225 int (*show_options)(struct seq_file *, struct vfsmount *);
227 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
228 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
231 All methods are called without any locks being held, unless otherwise
232 noted. This means that most methods can block safely. All methods are
233 only called from a process context (i.e. not from an interrupt handler
236 alloc_inode: this method is called by inode_alloc() to allocate memory
237 for struct inode and initialize it. If this function is not
238 defined, a simple 'struct inode' is allocated. Normally
239 alloc_inode will be used to allocate a larger structure which
240 contains a 'struct inode' embedded within it.
242 destroy_inode: this method is called by destroy_inode() to release
243 resources allocated for struct inode. It is only required if
244 ->alloc_inode was defined and simply undoes anything done by
247 read_inode: this method is called to read a specific inode from the
248 mounted filesystem. The i_ino member in the struct inode is
249 initialized by the VFS to indicate which inode to read. Other
250 members are filled in by this method.
252 You can set this to NULL and use iget5_locked() instead of iget()
253 to read inodes. This is necessary for filesystems for which the
254 inode number is not sufficient to identify an inode.
256 dirty_inode: this method is called by the VFS to mark an inode dirty.
258 write_inode: this method is called when the VFS needs to write an
259 inode to disc. The second parameter indicates whether the write
260 should be synchronous or not, not all filesystems check this flag.
262 put_inode: called when the VFS inode is removed from the inode
265 drop_inode: called when the last access to the inode is dropped,
266 with the inode_lock spinlock held.
268 This method should be either NULL (normal UNIX filesystem
269 semantics) or "generic_delete_inode" (for filesystems that do not
270 want to cache inodes - causing "delete_inode" to always be
271 called regardless of the value of i_nlink)
273 The "generic_delete_inode()" behavior is equivalent to the
274 old practice of using "force_delete" in the put_inode() case,
275 but does not have the races that the "force_delete()" approach
278 delete_inode: called when the VFS wants to delete an inode
280 put_super: called when the VFS wishes to free the superblock
281 (i.e. unmount). This is called with the superblock lock held
283 write_super: called when the VFS superblock needs to be written to
284 disc. This method is optional
286 sync_fs: called when VFS is writing out all dirty data associated with
287 a superblock. The second parameter indicates whether the method
288 should wait until the write out has been completed. Optional.
290 write_super_lockfs: called when VFS is locking a filesystem and
291 forcing it into a consistent state. This method is currently
292 used by the Logical Volume Manager (LVM).
294 unlockfs: called when VFS is unlocking a filesystem and making it writable
297 statfs: called when the VFS needs to get filesystem statistics. This
298 is called with the kernel lock held
300 remount_fs: called when the filesystem is remounted. This is called
301 with the kernel lock held
303 clear_inode: called then the VFS clears the inode. Optional
305 umount_begin: called when the VFS is unmounting a filesystem.
307 sync_inodes: called when the VFS is writing out dirty data associated with
310 show_options: called by the VFS to show mount options for /proc/<pid>/mounts.
312 quota_read: called by the VFS to read from filesystem quota file.
314 quota_write: called by the VFS to write to filesystem quota file.
316 The read_inode() method is responsible for filling in the "i_op"
317 field. This is a pointer to a "struct inode_operations" which
318 describes the methods that can be performed on individual inodes.
324 An inode object represents an object within the filesystem.
327 struct inode_operations
328 -----------------------
330 This describes how the VFS can manipulate an inode in your
331 filesystem. As of kernel 2.6.13, the following members are defined:
333 struct inode_operations {
334 int (*create) (struct inode *,struct dentry *,int, struct nameidata *);
335 struct dentry * (*lookup) (struct inode *,struct dentry *, struct nameidata *);
336 int (*link) (struct dentry *,struct inode *,struct dentry *);
337 int (*unlink) (struct inode *,struct dentry *);
338 int (*symlink) (struct inode *,struct dentry *,const char *);
339 int (*mkdir) (struct inode *,struct dentry *,int);
340 int (*rmdir) (struct inode *,struct dentry *);
341 int (*mknod) (struct inode *,struct dentry *,int,dev_t);
342 int (*rename) (struct inode *, struct dentry *,
343 struct inode *, struct dentry *);
344 int (*readlink) (struct dentry *, char __user *,int);
345 void * (*follow_link) (struct dentry *, struct nameidata *);
346 void (*put_link) (struct dentry *, struct nameidata *, void *);
347 void (*truncate) (struct inode *);
348 int (*permission) (struct inode *, int, struct nameidata *);
349 int (*setattr) (struct dentry *, struct iattr *);
350 int (*getattr) (struct vfsmount *mnt, struct dentry *, struct kstat *);
351 int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
352 ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t);
353 ssize_t (*listxattr) (struct dentry *, char *, size_t);
354 int (*removexattr) (struct dentry *, const char *);
357 Again, all methods are called without any locks being held, unless
360 create: called by the open(2) and creat(2) system calls. Only
361 required if you want to support regular files. The dentry you
362 get should not have an inode (i.e. it should be a negative
363 dentry). Here you will probably call d_instantiate() with the
364 dentry and the newly created inode
366 lookup: called when the VFS needs to look up an inode in a parent
367 directory. The name to look for is found in the dentry. This
368 method must call d_add() to insert the found inode into the
369 dentry. The "i_count" field in the inode structure should be
370 incremented. If the named inode does not exist a NULL inode
371 should be inserted into the dentry (this is called a negative
372 dentry). Returning an error code from this routine must only
373 be done on a real error, otherwise creating inodes with system
374 calls like create(2), mknod(2), mkdir(2) and so on will fail.
375 If you wish to overload the dentry methods then you should
376 initialise the "d_dop" field in the dentry; this is a pointer
377 to a struct "dentry_operations".
378 This method is called with the directory inode semaphore held
380 link: called by the link(2) system call. Only required if you want
381 to support hard links. You will probably need to call
382 d_instantiate() just as you would in the create() method
384 unlink: called by the unlink(2) system call. Only required if you
385 want to support deleting inodes
387 symlink: called by the symlink(2) system call. Only required if you
388 want to support symlinks. You will probably need to call
389 d_instantiate() just as you would in the create() method
391 mkdir: called by the mkdir(2) system call. Only required if you want
392 to support creating subdirectories. You will probably need to
393 call d_instantiate() just as you would in the create() method
395 rmdir: called by the rmdir(2) system call. Only required if you want
396 to support deleting subdirectories
398 mknod: called by the mknod(2) system call to create a device (char,
399 block) inode or a named pipe (FIFO) or socket. Only required
400 if you want to support creating these types of inodes. You
401 will probably need to call d_instantiate() just as you would
402 in the create() method
404 rename: called by the rename(2) system call to rename the object to
405 have the parent and name given by the second inode and dentry.
407 readlink: called by the readlink(2) system call. Only required if
408 you want to support reading symbolic links
410 follow_link: called by the VFS to follow a symbolic link to the
411 inode it points to. Only required if you want to support
412 symbolic links. This method returns a void pointer cookie
413 that is passed to put_link().
415 put_link: called by the VFS to release resources allocated by
416 follow_link(). The cookie returned by follow_link() is passed
417 to this method as the last parameter. It is used by
418 filesystems such as NFS where page cache is not stable
419 (i.e. page that was installed when the symbolic link walk
420 started might not be in the page cache at the end of the
423 truncate: called by the VFS to change the size of a file. The
424 i_size field of the inode is set to the desired size by the
425 VFS before this method is called. This method is called by
426 the truncate(2) system call and related functionality.
428 permission: called by the VFS to check for access rights on a POSIX-like
431 setattr: called by the VFS to set attributes for a file. This method
432 is called by chmod(2) and related system calls.
434 getattr: called by the VFS to get attributes of a file. This method
435 is called by stat(2) and related system calls.
437 setxattr: called by the VFS to set an extended attribute for a file.
438 Extended attribute is a name:value pair associated with an
439 inode. This method is called by setxattr(2) system call.
441 getxattr: called by the VFS to retrieve the value of an extended
442 attribute name. This method is called by getxattr(2) function
445 listxattr: called by the VFS to list all extended attributes for a
446 given file. This method is called by listxattr(2) system call.
448 removexattr: called by the VFS to remove an extended attribute from
449 a file. This method is called by removexattr(2) system call.
452 The Address Space Object
453 ========================
455 The address space object is used to group and manage pages in the page
456 cache. It can be used to keep track of the pages in a file (or
457 anything else) and also track the mapping of sections of the file into
458 process address spaces.
460 There are a number of distinct yet related services that an
461 address-space can provide. These include communicating memory
462 pressure, page lookup by address, and keeping track of pages tagged as
465 The first can be used independently to the others. The VM can try to
466 either write dirty pages in order to clean them, or release clean
467 pages in order to reuse them. To do this it can call the ->writepage
468 method on dirty pages, and ->releasepage on clean pages with
469 PagePrivate set. Clean pages without PagePrivate and with no external
470 references will be released without notice being given to the
473 To achieve this functionality, pages need to be placed on an LRU with
474 lru_cache_add and mark_page_active needs to be called whenever the
477 Pages are normally kept in a radix tree index by ->index. This tree
478 maintains information about the PG_Dirty and PG_Writeback status of
479 each page, so that pages with either of these flags can be found
482 The Dirty tag is primarily used by mpage_writepages - the default
483 ->writepages method. It uses the tag to find dirty pages to call
484 ->writepage on. If mpage_writepages is not used (i.e. the address
485 provides its own ->writepages) , the PAGECACHE_TAG_DIRTY tag is
486 almost unused. write_inode_now and sync_inode do use it (through
487 __sync_single_inode) to check if ->writepages has been successful in
488 writing out the whole address_space.
490 The Writeback tag is used by filemap*wait* and sync_page* functions,
491 via wait_on_page_writeback_range, to wait for all writeback to
492 complete. While waiting ->sync_page (if defined) will be called on
493 each page that is found to require writeback.
495 An address_space handler may attach extra information to a page,
496 typically using the 'private' field in the 'struct page'. If such
497 information is attached, the PG_Private flag should be set. This will
498 cause various VM routines to make extra calls into the address_space
499 handler to deal with that data.
501 An address space acts as an intermediate between storage and
502 application. Data is read into the address space a whole page at a
503 time, and provided to the application either by copying of the page,
504 or by memory-mapping the page.
505 Data is written into the address space by the application, and then
506 written-back to storage typically in whole pages, however the
507 address_space has finer control of write sizes.
509 The read process essentially only requires 'readpage'. The write
510 process is more complicated and uses prepare_write/commit_write or
511 set_page_dirty to write data into the address_space, and writepage,
512 sync_page, and writepages to writeback data to storage.
514 Adding and removing pages to/from an address_space is protected by the
517 When data is written to a page, the PG_Dirty flag should be set. It
518 typically remains set until writepage asks for it to be written. This
519 should clear PG_Dirty and set PG_Writeback. It can be actually
520 written at any point after PG_Dirty is clear. Once it is known to be
521 safe, PG_Writeback is cleared.
523 Writeback makes use of a writeback_control structure...
525 struct address_space_operations
526 -------------------------------
528 This describes how the VFS can manipulate mapping of a file to page cache in
529 your filesystem. As of kernel 2.6.16, the following members are defined:
531 struct address_space_operations {
532 int (*writepage)(struct page *page, struct writeback_control *wbc);
533 int (*readpage)(struct file *, struct page *);
534 int (*sync_page)(struct page *);
535 int (*writepages)(struct address_space *, struct writeback_control *);
536 int (*set_page_dirty)(struct page *page);
537 int (*readpages)(struct file *filp, struct address_space *mapping,
538 struct list_head *pages, unsigned nr_pages);
539 int (*prepare_write)(struct file *, struct page *, unsigned, unsigned);
540 int (*commit_write)(struct file *, struct page *, unsigned, unsigned);
541 sector_t (*bmap)(struct address_space *, sector_t);
542 int (*invalidatepage) (struct page *, unsigned long);
543 int (*releasepage) (struct page *, int);
544 ssize_t (*direct_IO)(int, struct kiocb *, const struct iovec *iov,
545 loff_t offset, unsigned long nr_segs);
546 struct page* (*get_xip_page)(struct address_space *, sector_t,
548 /* migrate the contents of a page to the specified target */
549 int (*migratepage) (struct page *, struct page *);
552 writepage: called by the VM to write a dirty page to backing store.
553 This may happen for data integrity reasons (i.e. 'sync'), or
554 to free up memory (flush). The difference can be seen in
556 The PG_Dirty flag has been cleared and PageLocked is true.
557 writepage should start writeout, should set PG_Writeback,
558 and should make sure the page is unlocked, either synchronously
559 or asynchronously when the write operation completes.
561 If wbc->sync_mode is WB_SYNC_NONE, ->writepage doesn't have to
562 try too hard if there are problems, and may choose to write out
563 other pages from the mapping if that is easier (e.g. due to
564 internal dependencies). If it chooses not to start writeout, it
565 should return AOP_WRITEPAGE_ACTIVATE so that the VM will not keep
566 calling ->writepage on that page.
568 See the file "Locking" for more details.
570 readpage: called by the VM to read a page from backing store.
571 The page will be Locked when readpage is called, and should be
572 unlocked and marked uptodate once the read completes.
573 If ->readpage discovers that it needs to unlock the page for
574 some reason, it can do so, and then return AOP_TRUNCATED_PAGE.
575 In this case, the page will be relocated, relocked and if
576 that all succeeds, ->readpage will be called again.
578 sync_page: called by the VM to notify the backing store to perform all
579 queued I/O operations for a page. I/O operations for other pages
580 associated with this address_space object may also be performed.
582 This function is optional and is called only for pages with
583 PG_Writeback set while waiting for the writeback to complete.
585 writepages: called by the VM to write out pages associated with the
586 address_space object. If wbc->sync_mode is WBC_SYNC_ALL, then
587 the writeback_control will specify a range of pages that must be
588 written out. If it is WBC_SYNC_NONE, then a nr_to_write is given
589 and that many pages should be written if possible.
590 If no ->writepages is given, then mpage_writepages is used
591 instead. This will choose pages from the address space that are
592 tagged as DIRTY and will pass them to ->writepage.
594 set_page_dirty: called by the VM to set a page dirty.
595 This is particularly needed if an address space attaches
596 private data to a page, and that data needs to be updated when
597 a page is dirtied. This is called, for example, when a memory
598 mapped page gets modified.
599 If defined, it should set the PageDirty flag, and the
600 PAGECACHE_TAG_DIRTY tag in the radix tree.
602 readpages: called by the VM to read pages associated with the address_space
603 object. This is essentially just a vector version of
604 readpage. Instead of just one page, several pages are
606 readpages is only used for read-ahead, so read errors are
607 ignored. If anything goes wrong, feel free to give up.
609 prepare_write: called by the generic write path in VM to set up a write
610 request for a page. This indicates to the address space that
611 the given range of bytes is about to be written. The
612 address_space should check that the write will be able to
613 complete, by allocating space if necessary and doing any other
614 internal housekeeping. If the write will update parts of
615 any basic-blocks on storage, then those blocks should be
616 pre-read (if they haven't been read already) so that the
617 updated blocks can be written out properly.
618 The page will be locked. If prepare_write wants to unlock the
619 page it, like readpage, may do so and return
621 In this case the prepare_write will be retried one the lock is
624 Note: the page _must not_ be marked uptodate in this function
625 (or anywhere else) unless it actually is uptodate right now. As
626 soon as a page is marked uptodate, it is possible for a concurrent
627 read(2) to copy it to userspace.
629 commit_write: If prepare_write succeeds, new data will be copied
630 into the page and then commit_write will be called. It will
631 typically update the size of the file (if appropriate) and
632 mark the inode as dirty, and do any other related housekeeping
633 operations. It should avoid returning an error if possible -
634 errors should have been handled by prepare_write.
636 bmap: called by the VFS to map a logical block offset within object to
637 physical block number. This method is used by the FIBMAP
638 ioctl and for working with swap-files. To be able to swap to
639 a file, the file must have a stable mapping to a block
640 device. The swap system does not go through the filesystem
641 but instead uses bmap to find out where the blocks in the file
642 are and uses those addresses directly.
645 invalidatepage: If a page has PagePrivate set, then invalidatepage
646 will be called when part or all of the page is to be removed
647 from the address space. This generally corresponds to either a
648 truncation or a complete invalidation of the address space
649 (in the latter case 'offset' will always be 0).
650 Any private data associated with the page should be updated
651 to reflect this truncation. If offset is 0, then
652 the private data should be released, because the page
653 must be able to be completely discarded. This may be done by
654 calling the ->releasepage function, but in this case the
655 release MUST succeed.
657 releasepage: releasepage is called on PagePrivate pages to indicate
658 that the page should be freed if possible. ->releasepage
659 should remove any private data from the page and clear the
660 PagePrivate flag. It may also remove the page from the
661 address_space. If this fails for some reason, it may indicate
662 failure with a 0 return value.
663 This is used in two distinct though related cases. The first
664 is when the VM finds a clean page with no active users and
665 wants to make it a free page. If ->releasepage succeeds, the
666 page will be removed from the address_space and become free.
668 The second case if when a request has been made to invalidate
669 some or all pages in an address_space. This can happen
670 through the fadvice(POSIX_FADV_DONTNEED) system call or by the
671 filesystem explicitly requesting it as nfs and 9fs do (when
672 they believe the cache may be out of date with storage) by
673 calling invalidate_inode_pages2().
674 If the filesystem makes such a call, and needs to be certain
675 that all pages are invalidated, then its releasepage will
676 need to ensure this. Possibly it can clear the PageUptodate
677 bit if it cannot free private data yet.
679 direct_IO: called by the generic read/write routines to perform
680 direct_IO - that is IO requests which bypass the page cache
681 and transfer data directly between the storage and the
682 application's address space.
684 get_xip_page: called by the VM to translate a block number to a page.
685 The page is valid until the corresponding filesystem is unmounted.
686 Filesystems that want to use execute-in-place (XIP) need to implement
687 it. An example implementation can be found in fs/ext2/xip.c.
689 migrate_page: This is used to compact the physical memory usage.
690 If the VM wants to relocate a page (maybe off a memory card
691 that is signalling imminent failure) it will pass a new page
692 and an old page to this function. migrate_page should
693 transfer any private data across and update any references
694 that it has to the page.
699 A file object represents a file opened by a process.
702 struct file_operations
703 ----------------------
705 This describes how the VFS can manipulate an open file. As of kernel
706 2.6.17, the following members are defined:
708 struct file_operations {
709 loff_t (*llseek) (struct file *, loff_t, int);
710 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
711 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
712 ssize_t (*aio_read) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
713 ssize_t (*aio_write) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
714 int (*readdir) (struct file *, void *, filldir_t);
715 unsigned int (*poll) (struct file *, struct poll_table_struct *);
716 int (*ioctl) (struct inode *, struct file *, unsigned int, unsigned long);
717 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
718 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
719 int (*mmap) (struct file *, struct vm_area_struct *);
720 int (*open) (struct inode *, struct file *);
721 int (*flush) (struct file *);
722 int (*release) (struct inode *, struct file *);
723 int (*fsync) (struct file *, struct dentry *, int datasync);
724 int (*aio_fsync) (struct kiocb *, int datasync);
725 int (*fasync) (int, struct file *, int);
726 int (*lock) (struct file *, int, struct file_lock *);
727 ssize_t (*readv) (struct file *, const struct iovec *, unsigned long, loff_t *);
728 ssize_t (*writev) (struct file *, const struct iovec *, unsigned long, loff_t *);
729 ssize_t (*sendfile) (struct file *, loff_t *, size_t, read_actor_t, void *);
730 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
731 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
732 int (*check_flags)(int);
733 int (*dir_notify)(struct file *filp, unsigned long arg);
734 int (*flock) (struct file *, int, struct file_lock *);
735 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, size_t, unsigned
737 ssize_t (*splice_read)(struct file *, struct pipe_inode_info *, size_t, unsigned
741 Again, all methods are called without any locks being held, unless
744 llseek: called when the VFS needs to move the file position index
746 read: called by read(2) and related system calls
748 aio_read: called by io_submit(2) and other asynchronous I/O operations
750 write: called by write(2) and related system calls
752 aio_write: called by io_submit(2) and other asynchronous I/O operations
754 readdir: called when the VFS needs to read the directory contents
756 poll: called by the VFS when a process wants to check if there is
757 activity on this file and (optionally) go to sleep until there
758 is activity. Called by the select(2) and poll(2) system calls
760 ioctl: called by the ioctl(2) system call
762 unlocked_ioctl: called by the ioctl(2) system call. Filesystems that do not
763 require the BKL should use this method instead of the ioctl() above.
765 compat_ioctl: called by the ioctl(2) system call when 32 bit system calls
766 are used on 64 bit kernels.
768 mmap: called by the mmap(2) system call
770 open: called by the VFS when an inode should be opened. When the VFS
771 opens a file, it creates a new "struct file". It then calls the
772 open method for the newly allocated file structure. You might
773 think that the open method really belongs in
774 "struct inode_operations", and you may be right. I think it's
775 done the way it is because it makes filesystems simpler to
776 implement. The open() method is a good place to initialize the
777 "private_data" member in the file structure if you want to point
778 to a device structure
780 flush: called by the close(2) system call to flush a file
782 release: called when the last reference to an open file is closed
784 fsync: called by the fsync(2) system call
786 fasync: called by the fcntl(2) system call when asynchronous
787 (non-blocking) mode is enabled for a file
789 lock: called by the fcntl(2) system call for F_GETLK, F_SETLK, and F_SETLKW
792 readv: called by the readv(2) system call
794 writev: called by the writev(2) system call
796 sendfile: called by the sendfile(2) system call
798 get_unmapped_area: called by the mmap(2) system call
800 check_flags: called by the fcntl(2) system call for F_SETFL command
802 dir_notify: called by the fcntl(2) system call for F_NOTIFY command
804 flock: called by the flock(2) system call
806 splice_write: called by the VFS to splice data from a pipe to a file. This
807 method is used by the splice(2) system call
809 splice_read: called by the VFS to splice data from file to a pipe. This
810 method is used by the splice(2) system call
812 Note that the file operations are implemented by the specific
813 filesystem in which the inode resides. When opening a device node
814 (character or block special) most filesystems will call special
815 support routines in the VFS which will locate the required device
816 driver information. These support routines replace the filesystem file
817 operations with those for the device driver, and then proceed to call
818 the new open() method for the file. This is how opening a device file
819 in the filesystem eventually ends up calling the device driver open()
823 Directory Entry Cache (dcache)
824 ==============================
827 struct dentry_operations
828 ------------------------
830 This describes how a filesystem can overload the standard dentry
831 operations. Dentries and the dcache are the domain of the VFS and the
832 individual filesystem implementations. Device drivers have no business
833 here. These methods may be set to NULL, as they are either optional or
834 the VFS uses a default. As of kernel 2.6.22, the following members are
837 struct dentry_operations {
838 int (*d_revalidate)(struct dentry *, struct nameidata *);
839 int (*d_hash) (struct dentry *, struct qstr *);
840 int (*d_compare) (struct dentry *, struct qstr *, struct qstr *);
841 int (*d_delete)(struct dentry *);
842 void (*d_release)(struct dentry *);
843 void (*d_iput)(struct dentry *, struct inode *);
844 char *(*d_dname)(struct dentry *, char *, int);
847 d_revalidate: called when the VFS needs to revalidate a dentry. This
848 is called whenever a name look-up finds a dentry in the
849 dcache. Most filesystems leave this as NULL, because all their
850 dentries in the dcache are valid
852 d_hash: called when the VFS adds a dentry to the hash table
854 d_compare: called when a dentry should be compared with another
856 d_delete: called when the last reference to a dentry is
857 deleted. This means no-one is using the dentry, however it is
858 still valid and in the dcache
860 d_release: called when a dentry is really deallocated
862 d_iput: called when a dentry loses its inode (just prior to its
863 being deallocated). The default when this is NULL is that the
864 VFS calls iput(). If you define this method, you must call
867 d_dname: called when the pathname of a dentry should be generated.
868 Usefull for some pseudo filesystems (sockfs, pipefs, ...) to delay
869 pathname generation. (Instead of doing it when dentry is created,
870 its done only when the path is needed.). Real filesystems probably
871 dont want to use it, because their dentries are present in global
872 dcache hash, so their hash should be an invariant. As no lock is
873 held, d_dname() should not try to modify the dentry itself, unless
874 appropriate SMP safety is used. CAUTION : d_path() logic is quite
875 tricky. The correct way to return for example "Hello" is to put it
876 at the end of the buffer, and returns a pointer to the first char.
877 dynamic_dname() helper function is provided to take care of this.
881 static char *pipefs_dname(struct dentry *dent, char *buffer, int buflen)
883 return dynamic_dname(dentry, buffer, buflen, "pipe:[%lu]",
884 dentry->d_inode->i_ino);
887 Each dentry has a pointer to its parent dentry, as well as a hash list
888 of child dentries. Child dentries are basically like files in a
892 Directory Entry Cache API
893 --------------------------
895 There are a number of functions defined which permit a filesystem to
898 dget: open a new handle for an existing dentry (this just increments
901 dput: close a handle for a dentry (decrements the usage count). If
902 the usage count drops to 0, the "d_delete" method is called
903 and the dentry is placed on the unused list if the dentry is
904 still in its parents hash list. Putting the dentry on the
905 unused list just means that if the system needs some RAM, it
906 goes through the unused list of dentries and deallocates them.
907 If the dentry has already been unhashed and the usage count
908 drops to 0, in this case the dentry is deallocated after the
909 "d_delete" method is called
911 d_drop: this unhashes a dentry from its parents hash list. A
912 subsequent call to dput() will deallocate the dentry if its
913 usage count drops to 0
915 d_delete: delete a dentry. If there are no other open references to
916 the dentry then the dentry is turned into a negative dentry
917 (the d_iput() method is called). If there are other
918 references, then d_drop() is called instead
920 d_add: add a dentry to its parents hash list and then calls
923 d_instantiate: add a dentry to the alias hash list for the inode and
924 updates the "d_inode" member. The "i_count" member in the
925 inode structure should be set/incremented. If the inode
926 pointer is NULL, the dentry is called a "negative
927 dentry". This function is commonly called when an inode is
928 created for an existing negative dentry
930 d_lookup: look up a dentry given its parent and path name component
931 It looks up the child of that given name from the dcache
932 hash table. If it is found, the reference count is incremented
933 and the dentry is returned. The caller must use d_put()
934 to free the dentry when it finishes using it.
936 For further information on dentry locking, please refer to the document
937 Documentation/filesystems/dentry-locking.txt.
943 (Note some of these resources are not up-to-date with the latest kernel
946 Creating Linux virtual filesystems. 2002
947 <http://lwn.net/Articles/13325/>
949 The Linux Virtual File-system Layer by Neil Brown. 1999
950 <http://www.cse.unsw.edu.au/~neilb/oss/linux-commentary/vfs.html>
952 A tour of the Linux VFS by Michael K. Johnson. 1996
953 <http://www.tldp.org/LDP/khg/HyperNews/get/fs/vfstour.html>
955 A small trail through the Linux kernel by Andries Brouwer. 2001
956 <http://www.win.tue.nl/~aeb/linux/vfs/trail.html>