1 2 Overview of the Linux Virtual File System 3 4 Original author: Richard Gooch <rgooch@atnf.csiro.au> 5 6 Last updated on June 24, 2007. 7 8 Copyright (C) 1999 Richard Gooch 9 Copyright (C) 2005 Pekka Enberg 10 11 This file is released under the GPLv2. 12 13 14Introduction 15============ 16 17The Virtual File System (also known as the Virtual Filesystem Switch) 18is the software layer in the kernel that provides the filesystem 19interface to userspace programs. It also provides an abstraction 20within the kernel which allows different filesystem implementations to 21coexist. 22 23VFS system calls open(2), stat(2), read(2), write(2), chmod(2) and so 24on are called from a process context. Filesystem locking is described 25in the document Documentation/filesystems/Locking. 26 27 28Directory Entry Cache (dcache) 29------------------------------ 30 31The VFS implements the open(2), stat(2), chmod(2), and similar system 32calls. The pathname argument that is passed to them is used by the VFS 33to search through the directory entry cache (also known as the dentry 34cache or dcache). This provides a very fast look-up mechanism to 35translate a pathname (filename) into a specific dentry. Dentries live 36in RAM and are never saved to disc: they exist only for performance. 37 38The dentry cache is meant to be a view into your entire filespace. As 39most computers cannot fit all dentries in the RAM at the same time, 40some bits of the cache are missing. In order to resolve your pathname 41into a dentry, the VFS may have to resort to creating dentries along 42the way, and then loading the inode. This is done by looking up the 43inode. 44 45 46The Inode Object 47---------------- 48 49An individual dentry usually has a pointer to an inode. Inodes are 50filesystem objects such as regular files, directories, FIFOs and other 51beasts. They live either on the disc (for block device filesystems) 52or in the memory (for pseudo filesystems). Inodes that live on the 53disc are copied into the memory when required and changes to the inode 54are written back to disc. A single inode can be pointed to by multiple 55dentries (hard links, for example, do this). 56 57To look up an inode requires that the VFS calls the lookup() method of 58the parent directory inode. This method is installed by the specific 59filesystem implementation that the inode lives in. Once the VFS has 60the required dentry (and hence the inode), we can do all those boring 61things like open(2) the file, or stat(2) it to peek at the inode 62data. The stat(2) operation is fairly simple: once the VFS has the 63dentry, it peeks at the inode data and passes some of it back to 64userspace. 65 66 67The File Object 68--------------- 69 70Opening a file requires another operation: allocation of a file 71structure (this is the kernel-side implementation of file 72descriptors). The freshly allocated file structure is initialized with 73a pointer to the dentry and a set of file operation member functions. 74These are taken from the inode data. The open() file method is then 75called so the specific filesystem implementation can do its work. You 76can see that this is another switch performed by the VFS. The file 77structure is placed into the file descriptor table for the process. 78 79Reading, writing and closing files (and other assorted VFS operations) 80is done by using the userspace file descriptor to grab the appropriate 81file structure, and then calling the required file structure method to 82do whatever is required. For as long as the file is open, it keeps the 83dentry in use, which in turn means that the VFS inode is still in use. 84 85 86Registering and Mounting a Filesystem 87===================================== 88 89To register and unregister a filesystem, use the following API 90functions: 91 92 #include <linux/fs.h> 93 94 extern int register_filesystem(struct file_system_type *); 95 extern int unregister_filesystem(struct file_system_type *); 96 97The passed struct file_system_type describes your filesystem. When a 98request is made to mount a filesystem onto a directory in your namespace, 99the VFS will call the appropriate mount() method for the specific 100filesystem. New vfsmount referring to the tree returned by ->mount() 101will be attached to the mountpoint, so that when pathname resolution 102reaches the mountpoint it will jump into the root of that vfsmount. 103 104You can see all filesystems that are registered to the kernel in the 105file /proc/filesystems. 106 107 108struct file_system_type 109----------------------- 110 111This describes the filesystem. As of kernel 2.6.39, the following 112members are defined: 113 114struct file_system_type { 115 const char *name; 116 int fs_flags; 117 struct dentry *(*mount) (struct file_system_type *, int, 118 const char *, void *); 119 void (*kill_sb) (struct super_block *); 120 struct module *owner; 121 struct file_system_type * next; 122 struct list_head fs_supers; 123 struct lock_class_key s_lock_key; 124 struct lock_class_key s_umount_key; 125}; 126 127 name: the name of the filesystem type, such as "ext2", "iso9660", 128 "msdos" and so on 129 130 fs_flags: various flags (i.e. FS_REQUIRES_DEV, FS_NO_DCACHE, etc.) 131 132 mount: the method to call when a new instance of this 133 filesystem should be mounted 134 135 kill_sb: the method to call when an instance of this filesystem 136 should be shut down 137 138 owner: for internal VFS use: you should initialize this to THIS_MODULE in 139 most cases. 140 141 next: for internal VFS use: you should initialize this to NULL 142 143 s_lock_key, s_umount_key: lockdep-specific 144 145The mount() method has the following arguments: 146 147 struct file_system_type *fs_type: describes the filesystem, partly initialized 148 by the specific filesystem code 149 150 int flags: mount flags 151 152 const char *dev_name: the device name we are mounting. 153 154 void *data: arbitrary mount options, usually comes as an ASCII 155 string (see "Mount Options" section) 156 157The mount() method must return the root dentry of the tree requested by 158caller. An active reference to its superblock must be grabbed and the 159superblock must be locked. On failure it should return ERR_PTR(error). 160 161The arguments match those of mount(2) and their interpretation 162depends on filesystem type. E.g. for block filesystems, dev_name is 163interpreted as block device name, that device is opened and if it 164contains a suitable filesystem image the method creates and initializes 165struct super_block accordingly, returning its root dentry to caller. 166 167->mount() may choose to return a subtree of existing filesystem - it 168doesn't have to create a new one. The main result from the caller's 169point of view is a reference to dentry at the root of (sub)tree to 170be attached; creation of new superblock is a common side effect. 171 172The most interesting member of the superblock structure that the 173mount() method fills in is the "s_op" field. This is a pointer to 174a "struct super_operations" which describes the next level of the 175filesystem implementation. 176 177Usually, a filesystem uses one of the generic mount() implementations 178and provides a fill_super() callback instead. The generic variants are: 179 180 mount_bdev: mount a filesystem residing on a block device 181 182 mount_nodev: mount a filesystem that is not backed by a device 183 184 mount_single: mount a filesystem which shares the instance between 185 all mounts 186 187A fill_super() callback implementation has the following arguments: 188 189 struct super_block *sb: the superblock structure. The callback 190 must initialize this properly. 191 192 void *data: arbitrary mount options, usually comes as an ASCII 193 string (see "Mount Options" section) 194 195 int silent: whether or not to be silent on error 196 197 198The Superblock Object 199===================== 200 201A superblock object represents a mounted filesystem. 202 203 204struct super_operations 205----------------------- 206 207This describes how the VFS can manipulate the superblock of your 208filesystem. As of kernel 2.6.22, the following members are defined: 209 210struct super_operations { 211 struct inode *(*alloc_inode)(struct super_block *sb); 212 void (*destroy_inode)(struct inode *); 213 214 void (*dirty_inode) (struct inode *, int flags); 215 int (*write_inode) (struct inode *, int); 216 void (*drop_inode) (struct inode *); 217 void (*delete_inode) (struct inode *); 218 void (*put_super) (struct super_block *); 219 int (*sync_fs)(struct super_block *sb, int wait); 220 int (*freeze_fs) (struct super_block *); 221 int (*unfreeze_fs) (struct super_block *); 222 int (*statfs) (struct dentry *, struct kstatfs *); 223 int (*remount_fs) (struct super_block *, int *, char *); 224 void (*clear_inode) (struct inode *); 225 void (*umount_begin) (struct super_block *); 226 227 int (*show_options)(struct seq_file *, struct dentry *); 228 229 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t); 230 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t); 231 int (*nr_cached_objects)(struct super_block *); 232 void (*free_cached_objects)(struct super_block *, int); 233}; 234 235All methods are called without any locks being held, unless otherwise 236noted. This means that most methods can block safely. All methods are 237only called from a process context (i.e. not from an interrupt handler 238or bottom half). 239 240 alloc_inode: this method is called by inode_alloc() to allocate memory 241 for struct inode and initialize it. If this function is not 242 defined, a simple 'struct inode' is allocated. Normally 243 alloc_inode will be used to allocate a larger structure which 244 contains a 'struct inode' embedded within it. 245 246 destroy_inode: this method is called by destroy_inode() to release 247 resources allocated for struct inode. It is only required if 248 ->alloc_inode was defined and simply undoes anything done by 249 ->alloc_inode. 250 251 dirty_inode: this method is called by the VFS to mark an inode dirty. 252 253 write_inode: this method is called when the VFS needs to write an 254 inode to disc. The second parameter indicates whether the write 255 should be synchronous or not, not all filesystems check this flag. 256 257 drop_inode: called when the last access to the inode is dropped, 258 with the inode->i_lock spinlock held. 259 260 This method should be either NULL (normal UNIX filesystem 261 semantics) or "generic_delete_inode" (for filesystems that do not 262 want to cache inodes - causing "delete_inode" to always be 263 called regardless of the value of i_nlink) 264 265 The "generic_delete_inode()" behavior is equivalent to the 266 old practice of using "force_delete" in the put_inode() case, 267 but does not have the races that the "force_delete()" approach 268 had. 269 270 delete_inode: called when the VFS wants to delete an inode 271 272 put_super: called when the VFS wishes to free the superblock 273 (i.e. unmount). This is called with the superblock lock held 274 275 sync_fs: called when VFS is writing out all dirty data associated with 276 a superblock. The second parameter indicates whether the method 277 should wait until the write out has been completed. Optional. 278 279 freeze_fs: called when VFS is locking a filesystem and 280 forcing it into a consistent state. This method is currently 281 used by the Logical Volume Manager (LVM). 282 283 unfreeze_fs: called when VFS is unlocking a filesystem and making it writable 284 again. 285 286 statfs: called when the VFS needs to get filesystem statistics. 287 288 remount_fs: called when the filesystem is remounted. This is called 289 with the kernel lock held 290 291 clear_inode: called then the VFS clears the inode. Optional 292 293 umount_begin: called when the VFS is unmounting a filesystem. 294 295 show_options: called by the VFS to show mount options for 296 /proc/<pid>/mounts. (see "Mount Options" section) 297 298 quota_read: called by the VFS to read from filesystem quota file. 299 300 quota_write: called by the VFS to write to filesystem quota file. 301 302 nr_cached_objects: called by the sb cache shrinking function for the 303 filesystem to return the number of freeable cached objects it contains. 304 Optional. 305 306 free_cache_objects: called by the sb cache shrinking function for the 307 filesystem to scan the number of objects indicated to try to free them. 308 Optional, but any filesystem implementing this method needs to also 309 implement ->nr_cached_objects for it to be called correctly. 310 311 We can't do anything with any errors that the filesystem might 312 encountered, hence the void return type. This will never be called if 313 the VM is trying to reclaim under GFP_NOFS conditions, hence this 314 method does not need to handle that situation itself. 315 316 Implementations must include conditional reschedule calls inside any 317 scanning loop that is done. This allows the VFS to determine 318 appropriate scan batch sizes without having to worry about whether 319 implementations will cause holdoff problems due to large scan batch 320 sizes. 321 322Whoever sets up the inode is responsible for filling in the "i_op" field. This 323is a pointer to a "struct inode_operations" which describes the methods that 324can be performed on individual inodes. 325 326 327The Inode Object 328================ 329 330An inode object represents an object within the filesystem. 331 332 333struct inode_operations 334----------------------- 335 336This describes how the VFS can manipulate an inode in your 337filesystem. As of kernel 2.6.22, the following members are defined: 338 339struct inode_operations { 340 int (*create) (struct inode *,struct dentry *, umode_t, bool); 341 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int); 342 int (*link) (struct dentry *,struct inode *,struct dentry *); 343 int (*unlink) (struct inode *,struct dentry *); 344 int (*symlink) (struct inode *,struct dentry *,const char *); 345 int (*mkdir) (struct inode *,struct dentry *,umode_t); 346 int (*rmdir) (struct inode *,struct dentry *); 347 int (*mknod) (struct inode *,struct dentry *,umode_t,dev_t); 348 int (*rename) (struct inode *, struct dentry *, 349 struct inode *, struct dentry *); 350 int (*readlink) (struct dentry *, char __user *,int); 351 void * (*follow_link) (struct dentry *, struct nameidata *); 352 void (*put_link) (struct dentry *, struct nameidata *, void *); 353 int (*permission) (struct inode *, int); 354 int (*get_acl)(struct inode *, int); 355 int (*setattr) (struct dentry *, struct iattr *); 356 int (*getattr) (struct vfsmount *mnt, struct dentry *, struct kstat *); 357 int (*setxattr) (struct dentry *, const char *,const void *,size_t,int); 358 ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t); 359 ssize_t (*listxattr) (struct dentry *, char *, size_t); 360 int (*removexattr) (struct dentry *, const char *); 361 void (*update_time)(struct inode *, struct timespec *, int); 362 int (*atomic_open)(struct inode *, struct dentry *, 363 struct file *, unsigned open_flag, 364 umode_t create_mode, int *opened); 365}; 366 367Again, all methods are called without any locks being held, unless 368otherwise noted. 369 370 create: called by the open(2) and creat(2) system calls. Only 371 required if you want to support regular files. The dentry you 372 get should not have an inode (i.e. it should be a negative 373 dentry). Here you will probably call d_instantiate() with the 374 dentry and the newly created inode 375 376 lookup: called when the VFS needs to look up an inode in a parent 377 directory. The name to look for is found in the dentry. This 378 method must call d_add() to insert the found inode into the 379 dentry. The "i_count" field in the inode structure should be 380 incremented. If the named inode does not exist a NULL inode 381 should be inserted into the dentry (this is called a negative 382 dentry). Returning an error code from this routine must only 383 be done on a real error, otherwise creating inodes with system 384 calls like create(2), mknod(2), mkdir(2) and so on will fail. 385 If you wish to overload the dentry methods then you should 386 initialise the "d_dop" field in the dentry; this is a pointer 387 to a struct "dentry_operations". 388 This method is called with the directory inode semaphore held 389 390 link: called by the link(2) system call. Only required if you want 391 to support hard links. You will probably need to call 392 d_instantiate() just as you would in the create() method 393 394 unlink: called by the unlink(2) system call. Only required if you 395 want to support deleting inodes 396 397 symlink: called by the symlink(2) system call. Only required if you 398 want to support symlinks. You will probably need to call 399 d_instantiate() just as you would in the create() method 400 401 mkdir: called by the mkdir(2) system call. Only required if you want 402 to support creating subdirectories. You will probably need to 403 call d_instantiate() just as you would in the create() method 404 405 rmdir: called by the rmdir(2) system call. Only required if you want 406 to support deleting subdirectories 407 408 mknod: called by the mknod(2) system call to create a device (char, 409 block) inode or a named pipe (FIFO) or socket. Only required 410 if you want to support creating these types of inodes. You 411 will probably need to call d_instantiate() just as you would 412 in the create() method 413 414 rename: called by the rename(2) system call to rename the object to 415 have the parent and name given by the second inode and dentry. 416 417 readlink: called by the readlink(2) system call. Only required if 418 you want to support reading symbolic links 419 420 follow_link: called by the VFS to follow a symbolic link to the 421 inode it points to. Only required if you want to support 422 symbolic links. This method returns a void pointer cookie 423 that is passed to put_link(). 424 425 put_link: called by the VFS to release resources allocated by 426 follow_link(). The cookie returned by follow_link() is passed 427 to this method as the last parameter. It is used by 428 filesystems such as NFS where page cache is not stable 429 (i.e. page that was installed when the symbolic link walk 430 started might not be in the page cache at the end of the 431 walk). 432 433 permission: called by the VFS to check for access rights on a POSIX-like 434 filesystem. 435 436 May be called in rcu-walk mode (mask & MAY_NOT_BLOCK). If in rcu-walk 437 mode, the filesystem must check the permission without blocking or 438 storing to the inode. 439 440 If a situation is encountered that rcu-walk cannot handle, return 441 -ECHILD and it will be called again in ref-walk mode. 442 443 setattr: called by the VFS to set attributes for a file. This method 444 is called by chmod(2) and related system calls. 445 446 getattr: called by the VFS to get attributes of a file. This method 447 is called by stat(2) and related system calls. 448 449 setxattr: called by the VFS to set an extended attribute for a file. 450 Extended attribute is a name:value pair associated with an 451 inode. This method is called by setxattr(2) system call. 452 453 getxattr: called by the VFS to retrieve the value of an extended 454 attribute name. This method is called by getxattr(2) function 455 call. 456 457 listxattr: called by the VFS to list all extended attributes for a 458 given file. This method is called by listxattr(2) system call. 459 460 removexattr: called by the VFS to remove an extended attribute from 461 a file. This method is called by removexattr(2) system call. 462 463 update_time: called by the VFS to update a specific time or the i_version of 464 an inode. If this is not defined the VFS will update the inode itself 465 and call mark_inode_dirty_sync. 466 467 atomic_open: called on the last component of an open. Using this optional 468 method the filesystem can look up, possibly create and open the file in 469 one atomic operation. If it cannot perform this (e.g. the file type 470 turned out to be wrong) it may signal this by returning 1 instead of 471 usual 0 or -ve . This method is only called if the last 472 component is negative or needs lookup. Cached positive dentries are 473 still handled by f_op->open(). 474 475The Address Space Object 476======================== 477 478The address space object is used to group and manage pages in the page 479cache. It can be used to keep track of the pages in a file (or 480anything else) and also track the mapping of sections of the file into 481process address spaces. 482 483There are a number of distinct yet related services that an 484address-space can provide. These include communicating memory 485pressure, page lookup by address, and keeping track of pages tagged as 486Dirty or Writeback. 487 488The first can be used independently to the others. The VM can try to 489either write dirty pages in order to clean them, or release clean 490pages in order to reuse them. To do this it can call the ->writepage 491method on dirty pages, and ->releasepage on clean pages with 492PagePrivate set. Clean pages without PagePrivate and with no external 493references will be released without notice being given to the 494address_space. 495 496To achieve this functionality, pages need to be placed on an LRU with 497lru_cache_add and mark_page_active needs to be called whenever the 498page is used. 499 500Pages are normally kept in a radix tree index by ->index. This tree 501maintains information about the PG_Dirty and PG_Writeback status of 502each page, so that pages with either of these flags can be found 503quickly. 504 505The Dirty tag is primarily used by mpage_writepages - the default 506->writepages method. It uses the tag to find dirty pages to call 507->writepage on. If mpage_writepages is not used (i.e. the address 508provides its own ->writepages) , the PAGECACHE_TAG_DIRTY tag is 509almost unused. write_inode_now and sync_inode do use it (through 510__sync_single_inode) to check if ->writepages has been successful in 511writing out the whole address_space. 512 513The Writeback tag is used by filemap*wait* and sync_page* functions, 514via filemap_fdatawait_range, to wait for all writeback to 515complete. While waiting ->sync_page (if defined) will be called on 516each page that is found to require writeback. 517 518An address_space handler may attach extra information to a page, 519typically using the 'private' field in the 'struct page'. If such 520information is attached, the PG_Private flag should be set. This will 521cause various VM routines to make extra calls into the address_space 522handler to deal with that data. 523 524An address space acts as an intermediate between storage and 525application. Data is read into the address space a whole page at a 526time, and provided to the application either by copying of the page, 527or by memory-mapping the page. 528Data is written into the address space by the application, and then 529written-back to storage typically in whole pages, however the 530address_space has finer control of write sizes. 531 532The read process essentially only requires 'readpage'. The write 533process is more complicated and uses write_begin/write_end or 534set_page_dirty to write data into the address_space, and writepage, 535sync_page, and writepages to writeback data to storage. 536 537Adding and removing pages to/from an address_space is protected by the 538inode's i_mutex. 539 540When data is written to a page, the PG_Dirty flag should be set. It 541typically remains set until writepage asks for it to be written. This 542should clear PG_Dirty and set PG_Writeback. It can be actually 543written at any point after PG_Dirty is clear. Once it is known to be 544safe, PG_Writeback is cleared. 545 546Writeback makes use of a writeback_control structure... 547 548struct address_space_operations 549------------------------------- 550 551This describes how the VFS can manipulate mapping of a file to page cache in 552your filesystem. The following members are defined: 553 554struct address_space_operations { 555 int (*writepage)(struct page *page, struct writeback_control *wbc); 556 int (*readpage)(struct file *, struct page *); 557 int (*sync_page)(struct page *); 558 int (*writepages)(struct address_space *, struct writeback_control *); 559 int (*set_page_dirty)(struct page *page); 560 int (*readpages)(struct file *filp, struct address_space *mapping, 561 struct list_head *pages, unsigned nr_pages); 562 int (*write_begin)(struct file *, struct address_space *mapping, 563 loff_t pos, unsigned len, unsigned flags, 564 struct page **pagep, void **fsdata); 565 int (*write_end)(struct file *, struct address_space *mapping, 566 loff_t pos, unsigned len, unsigned copied, 567 struct page *page, void *fsdata); 568 sector_t (*bmap)(struct address_space *, sector_t); 569 void (*invalidatepage) (struct page *, unsigned int, unsigned int); 570 int (*releasepage) (struct page *, int); 571 void (*freepage)(struct page *); 572 ssize_t (*direct_IO)(int, struct kiocb *, const struct iovec *iov, 573 loff_t offset, unsigned long nr_segs); 574 struct page* (*get_xip_page)(struct address_space *, sector_t, 575 int); 576 /* migrate the contents of a page to the specified target */ 577 int (*migratepage) (struct page *, struct page *); 578 int (*launder_page) (struct page *); 579 int (*error_remove_page) (struct mapping *mapping, struct page *page); 580 int (*swap_activate)(struct file *); 581 int (*swap_deactivate)(struct file *); 582}; 583 584 writepage: called by the VM to write a dirty page to backing store. 585 This may happen for data integrity reasons (i.e. 'sync'), or 586 to free up memory (flush). The difference can be seen in 587 wbc->sync_mode. 588 The PG_Dirty flag has been cleared and PageLocked is true. 589 writepage should start writeout, should set PG_Writeback, 590 and should make sure the page is unlocked, either synchronously 591 or asynchronously when the write operation completes. 592 593 If wbc->sync_mode is WB_SYNC_NONE, ->writepage doesn't have to 594 try too hard if there are problems, and may choose to write out 595 other pages from the mapping if that is easier (e.g. due to 596 internal dependencies). If it chooses not to start writeout, it 597 should return AOP_WRITEPAGE_ACTIVATE so that the VM will not keep 598 calling ->writepage on that page. 599 600 See the file "Locking" for more details. 601 602 readpage: called by the VM to read a page from backing store. 603 The page will be Locked when readpage is called, and should be 604 unlocked and marked uptodate once the read completes. 605 If ->readpage discovers that it needs to unlock the page for 606 some reason, it can do so, and then return AOP_TRUNCATED_PAGE. 607 In this case, the page will be relocated, relocked and if 608 that all succeeds, ->readpage will be called again. 609 610 sync_page: called by the VM to notify the backing store to perform all 611 queued I/O operations for a page. I/O operations for other pages 612 associated with this address_space object may also be performed. 613 614 This function is optional and is called only for pages with 615 PG_Writeback set while waiting for the writeback to complete. 616 617 writepages: called by the VM to write out pages associated with the 618 address_space object. If wbc->sync_mode is WBC_SYNC_ALL, then 619 the writeback_control will specify a range of pages that must be 620 written out. If it is WBC_SYNC_NONE, then a nr_to_write is given 621 and that many pages should be written if possible. 622 If no ->writepages is given, then mpage_writepages is used 623 instead. This will choose pages from the address space that are 624 tagged as DIRTY and will pass them to ->writepage. 625 626 set_page_dirty: called by the VM to set a page dirty. 627 This is particularly needed if an address space attaches 628 private data to a page, and that data needs to be updated when 629 a page is dirtied. This is called, for example, when a memory 630 mapped page gets modified. 631 If defined, it should set the PageDirty flag, and the 632 PAGECACHE_TAG_DIRTY tag in the radix tree. 633 634 readpages: called by the VM to read pages associated with the address_space 635 object. This is essentially just a vector version of 636 readpage. Instead of just one page, several pages are 637 requested. 638 readpages is only used for read-ahead, so read errors are 639 ignored. If anything goes wrong, feel free to give up. 640 641 write_begin: 642 Called by the generic buffered write code to ask the filesystem to 643 prepare to write len bytes at the given offset in the file. The 644 address_space should check that the write will be able to complete, 645 by allocating space if necessary and doing any other internal 646 housekeeping. If the write will update parts of any basic-blocks on 647 storage, then those blocks should be pre-read (if they haven't been 648 read already) so that the updated blocks can be written out properly. 649 650 The filesystem must return the locked pagecache page for the specified 651 offset, in *pagep, for the caller to write into. 652 653 It must be able to cope with short writes (where the length passed to 654 write_begin is greater than the number of bytes copied into the page). 655 656 flags is a field for AOP_FLAG_xxx flags, described in 657 include/linux/fs.h. 658 659 A void * may be returned in fsdata, which then gets passed into 660 write_end. 661 662 Returns 0 on success; < 0 on failure (which is the error code), in 663 which case write_end is not called. 664 665 write_end: After a successful write_begin, and data copy, write_end must 666 be called. len is the original len passed to write_begin, and copied 667 is the amount that was able to be copied (copied == len is always true 668 if write_begin was called with the AOP_FLAG_UNINTERRUPTIBLE flag). 669 670 The filesystem must take care of unlocking the page and releasing it 671 refcount, and updating i_size. 672 673 Returns < 0 on failure, otherwise the number of bytes (<= 'copied') 674 that were able to be copied into pagecache. 675 676 bmap: called by the VFS to map a logical block offset within object to 677 physical block number. This method is used by the FIBMAP 678 ioctl and for working with swap-files. To be able to swap to 679 a file, the file must have a stable mapping to a block 680 device. The swap system does not go through the filesystem 681 but instead uses bmap to find out where the blocks in the file 682 are and uses those addresses directly. 683 684 685 invalidatepage: If a page has PagePrivate set, then invalidatepage 686 will be called when part or all of the page is to be removed 687 from the address space. This generally corresponds to either a 688 truncation, punch hole or a complete invalidation of the address 689 space (in the latter case 'offset' will always be 0 and 'length' 690 will be PAGE_CACHE_SIZE). Any private data associated with the page 691 should be updated to reflect this truncation. If offset is 0 and 692 length is PAGE_CACHE_SIZE, then the private data should be released, 693 because the page must be able to be completely discarded. This may 694 be done by calling the ->releasepage function, but in this case the 695 release MUST succeed. 696 697 releasepage: releasepage is called on PagePrivate pages to indicate 698 that the page should be freed if possible. ->releasepage 699 should remove any private data from the page and clear the 700 PagePrivate flag. If releasepage() fails for some reason, it must 701 indicate failure with a 0 return value. 702 releasepage() is used in two distinct though related cases. The 703 first is when the VM finds a clean page with no active users and 704 wants to make it a free page. If ->releasepage succeeds, the 705 page will be removed from the address_space and become free. 706 707 The second case is when a request has been made to invalidate 708 some or all pages in an address_space. This can happen 709 through the fadvice(POSIX_FADV_DONTNEED) system call or by the 710 filesystem explicitly requesting it as nfs and 9fs do (when 711 they believe the cache may be out of date with storage) by 712 calling invalidate_inode_pages2(). 713 If the filesystem makes such a call, and needs to be certain 714 that all pages are invalidated, then its releasepage will 715 need to ensure this. Possibly it can clear the PageUptodate 716 bit if it cannot free private data yet. 717 718 freepage: freepage is called once the page is no longer visible in 719 the page cache in order to allow the cleanup of any private 720 data. Since it may be called by the memory reclaimer, it 721 should not assume that the original address_space mapping still 722 exists, and it should not block. 723 724 direct_IO: called by the generic read/write routines to perform 725 direct_IO - that is IO requests which bypass the page cache 726 and transfer data directly between the storage and the 727 application's address space. 728 729 get_xip_page: called by the VM to translate a block number to a page. 730 The page is valid until the corresponding filesystem is unmounted. 731 Filesystems that want to use execute-in-place (XIP) need to implement 732 it. An example implementation can be found in fs/ext2/xip.c. 733 734 migrate_page: This is used to compact the physical memory usage. 735 If the VM wants to relocate a page (maybe off a memory card 736 that is signalling imminent failure) it will pass a new page 737 and an old page to this function. migrate_page should 738 transfer any private data across and update any references 739 that it has to the page. 740 741 launder_page: Called before freeing a page - it writes back the dirty page. To 742 prevent redirtying the page, it is kept locked during the whole 743 operation. 744 745 error_remove_page: normally set to generic_error_remove_page if truncation 746 is ok for this address space. Used for memory failure handling. 747 Setting this implies you deal with pages going away under you, 748 unless you have them locked or reference counts increased. 749 750 swap_activate: Called when swapon is used on a file to allocate 751 space if necessary and pin the block lookup information in 752 memory. A return value of zero indicates success, 753 in which case this file can be used to back swapspace. The 754 swapspace operations will be proxied to this address space's 755 ->swap_{out,in} methods. 756 757 swap_deactivate: Called during swapoff on files where swap_activate 758 was successful. 759 760 761The File Object 762=============== 763 764A file object represents a file opened by a process. 765 766 767struct file_operations 768---------------------- 769 770This describes how the VFS can manipulate an open file. As of kernel 7713.5, the following members are defined: 772 773struct file_operations { 774 struct module *owner; 775 loff_t (*llseek) (struct file *, loff_t, int); 776 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *); 777 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *); 778 ssize_t (*aio_read) (struct kiocb *, const struct iovec *, unsigned long, loff_t); 779 ssize_t (*aio_write) (struct kiocb *, const struct iovec *, unsigned long, loff_t); 780 int (*readdir) (struct file *, void *, filldir_t); 781 unsigned int (*poll) (struct file *, struct poll_table_struct *); 782 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long); 783 long (*compat_ioctl) (struct file *, unsigned int, unsigned long); 784 int (*mmap) (struct file *, struct vm_area_struct *); 785 int (*open) (struct inode *, struct file *); 786 int (*flush) (struct file *); 787 int (*release) (struct inode *, struct file *); 788 int (*fsync) (struct file *, loff_t, loff_t, int datasync); 789 int (*aio_fsync) (struct kiocb *, int datasync); 790 int (*fasync) (int, struct file *, int); 791 int (*lock) (struct file *, int, struct file_lock *); 792 ssize_t (*readv) (struct file *, const struct iovec *, unsigned long, loff_t *); 793 ssize_t (*writev) (struct file *, const struct iovec *, unsigned long, loff_t *); 794 ssize_t (*sendfile) (struct file *, loff_t *, size_t, read_actor_t, void *); 795 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int); 796 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 797 int (*check_flags)(int); 798 int (*flock) (struct file *, int, struct file_lock *); 799 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, size_t, unsigned int); 800 ssize_t (*splice_read)(struct file *, struct pipe_inode_info *, size_t, unsigned int); 801 int (*setlease)(struct file *, long arg, struct file_lock **); 802 long (*fallocate)(struct file *, int mode, loff_t offset, loff_t len); 803}; 804 805Again, all methods are called without any locks being held, unless 806otherwise noted. 807 808 llseek: called when the VFS needs to move the file position index 809 810 read: called by read(2) and related system calls 811 812 aio_read: called by io_submit(2) and other asynchronous I/O operations 813 814 write: called by write(2) and related system calls 815 816 aio_write: called by io_submit(2) and other asynchronous I/O operations 817 818 readdir: called when the VFS needs to read the directory contents 819 820 poll: called by the VFS when a process wants to check if there is 821 activity on this file and (optionally) go to sleep until there 822 is activity. Called by the select(2) and poll(2) system calls 823 824 unlocked_ioctl: called by the ioctl(2) system call. 825 826 compat_ioctl: called by the ioctl(2) system call when 32 bit system calls 827 are used on 64 bit kernels. 828 829 mmap: called by the mmap(2) system call 830 831 open: called by the VFS when an inode should be opened. When the VFS 832 opens a file, it creates a new "struct file". It then calls the 833 open method for the newly allocated file structure. You might 834 think that the open method really belongs in 835 "struct inode_operations", and you may be right. I think it's 836 done the way it is because it makes filesystems simpler to 837 implement. The open() method is a good place to initialize the 838 "private_data" member in the file structure if you want to point 839 to a device structure 840 841 flush: called by the close(2) system call to flush a file 842 843 release: called when the last reference to an open file is closed 844 845 fsync: called by the fsync(2) system call 846 847 fasync: called by the fcntl(2) system call when asynchronous 848 (non-blocking) mode is enabled for a file 849 850 lock: called by the fcntl(2) system call for F_GETLK, F_SETLK, and F_SETLKW 851 commands 852 853 readv: called by the readv(2) system call 854 855 writev: called by the writev(2) system call 856 857 sendfile: called by the sendfile(2) system call 858 859 get_unmapped_area: called by the mmap(2) system call 860 861 check_flags: called by the fcntl(2) system call for F_SETFL command 862 863 flock: called by the flock(2) system call 864 865 splice_write: called by the VFS to splice data from a pipe to a file. This 866 method is used by the splice(2) system call 867 868 splice_read: called by the VFS to splice data from file to a pipe. This 869 method is used by the splice(2) system call 870 871 setlease: called by the VFS to set or release a file lock lease. 872 setlease has the file_lock_lock held and must not sleep. 873 874 fallocate: called by the VFS to preallocate blocks or punch a hole. 875 876Note that the file operations are implemented by the specific 877filesystem in which the inode resides. When opening a device node 878(character or block special) most filesystems will call special 879support routines in the VFS which will locate the required device 880driver information. These support routines replace the filesystem file 881operations with those for the device driver, and then proceed to call 882the new open() method for the file. This is how opening a device file 883in the filesystem eventually ends up calling the device driver open() 884method. 885 886 887Directory Entry Cache (dcache) 888============================== 889 890 891struct dentry_operations 892------------------------ 893 894This describes how a filesystem can overload the standard dentry 895operations. Dentries and the dcache are the domain of the VFS and the 896individual filesystem implementations. Device drivers have no business 897here. These methods may be set to NULL, as they are either optional or 898the VFS uses a default. As of kernel 2.6.22, the following members are 899defined: 900 901struct dentry_operations { 902 int (*d_revalidate)(struct dentry *, unsigned int); 903 int (*d_weak_revalidate)(struct dentry *, unsigned int); 904 int (*d_hash)(const struct dentry *, const struct inode *, 905 struct qstr *); 906 int (*d_compare)(const struct dentry *, const struct inode *, 907 const struct dentry *, const struct inode *, 908 unsigned int, const char *, const struct qstr *); 909 int (*d_delete)(const struct dentry *); 910 void (*d_release)(struct dentry *); 911 void (*d_iput)(struct dentry *, struct inode *); 912 char *(*d_dname)(struct dentry *, char *, int); 913 struct vfsmount *(*d_automount)(struct path *); 914 int (*d_manage)(struct dentry *, bool); 915}; 916 917 d_revalidate: called when the VFS needs to revalidate a dentry. This 918 is called whenever a name look-up finds a dentry in the 919 dcache. Most local filesystems leave this as NULL, because all their 920 dentries in the dcache are valid. Network filesystems are different 921 since things can change on the server without the client necessarily 922 being aware of it. 923 924 This function should return a positive value if the dentry is still 925 valid, and zero or a negative error code if it isn't. 926 927 d_revalidate may be called in rcu-walk mode (flags & LOOKUP_RCU). 928 If in rcu-walk mode, the filesystem must revalidate the dentry without 929 blocking or storing to the dentry, d_parent and d_inode should not be 930 used without care (because they can change and, in d_inode case, even 931 become NULL under us). 932 933 If a situation is encountered that rcu-walk cannot handle, return 934 -ECHILD and it will be called again in ref-walk mode. 935 936 d_weak_revalidate: called when the VFS needs to revalidate a "jumped" dentry. 937 This is called when a path-walk ends at dentry that was not acquired by 938 doing a lookup in the parent directory. This includes "/", "." and "..", 939 as well as procfs-style symlinks and mountpoint traversal. 940 941 In this case, we are less concerned with whether the dentry is still 942 fully correct, but rather that the inode is still valid. As with 943 d_revalidate, most local filesystems will set this to NULL since their 944 dcache entries are always valid. 945 946 This function has the same return code semantics as d_revalidate. 947 948 d_weak_revalidate is only called after leaving rcu-walk mode. 949 950 d_hash: called when the VFS adds a dentry to the hash table. The first 951 dentry passed to d_hash is the parent directory that the name is 952 to be hashed into. The inode is the dentry's inode. 953 954 Same locking and synchronisation rules as d_compare regarding 955 what is safe to dereference etc. 956 957 d_compare: called to compare a dentry name with a given name. The first 958 dentry is the parent of the dentry to be compared, the second is 959 the parent's inode, then the dentry and inode (may be NULL) of the 960 child dentry. len and name string are properties of the dentry to be 961 compared. qstr is the name to compare it with. 962 963 Must be constant and idempotent, and should not take locks if 964 possible, and should not or store into the dentry or inodes. 965 Should not dereference pointers outside the dentry or inodes without 966 lots of care (eg. d_parent, d_inode, d_name should not be used). 967 968 However, our vfsmount is pinned, and RCU held, so the dentries and 969 inodes won't disappear, neither will our sb or filesystem module. 970 ->i_sb and ->d_sb may be used. 971 972 It is a tricky calling convention because it needs to be called under 973 "rcu-walk", ie. without any locks or references on things. 974 975 d_delete: called when the last reference to a dentry is dropped and the 976 dcache is deciding whether or not to cache it. Return 1 to delete 977 immediately, or 0 to cache the dentry. Default is NULL which means to 978 always cache a reachable dentry. d_delete must be constant and 979 idempotent. 980 981 d_release: called when a dentry is really deallocated 982 983 d_iput: called when a dentry loses its inode (just prior to its 984 being deallocated). The default when this is NULL is that the 985 VFS calls iput(). If you define this method, you must call 986 iput() yourself 987 988 d_dname: called when the pathname of a dentry should be generated. 989 Useful for some pseudo filesystems (sockfs, pipefs, ...) to delay 990 pathname generation. (Instead of doing it when dentry is created, 991 it's done only when the path is needed.). Real filesystems probably 992 dont want to use it, because their dentries are present in global 993 dcache hash, so their hash should be an invariant. As no lock is 994 held, d_dname() should not try to modify the dentry itself, unless 995 appropriate SMP safety is used. CAUTION : d_path() logic is quite 996 tricky. The correct way to return for example "Hello" is to put it 997 at the end of the buffer, and returns a pointer to the first char. 998 dynamic_dname() helper function is provided to take care of this. 999 1000 d_automount: called when an automount dentry is to be traversed (optional). 1001 This should create a new VFS mount record and return the record to the 1002 caller. The caller is supplied with a path parameter giving the 1003 automount directory to describe the automount target and the parent 1004 VFS mount record to provide inheritable mount parameters. NULL should 1005 be returned if someone else managed to make the automount first. If 1006 the vfsmount creation failed, then an error code should be returned. 1007 If -EISDIR is returned, then the directory will be treated as an 1008 ordinary directory and returned to pathwalk to continue walking. 1009 1010 If a vfsmount is returned, the caller will attempt to mount it on the 1011 mountpoint and will remove the vfsmount from its expiration list in 1012 the case of failure. The vfsmount should be returned with 2 refs on 1013 it to prevent automatic expiration - the caller will clean up the 1014 additional ref. 1015 1016 This function is only used if DCACHE_NEED_AUTOMOUNT is set on the 1017 dentry. This is set by __d_instantiate() if S_AUTOMOUNT is set on the 1018 inode being added. 1019 1020 d_manage: called to allow the filesystem to manage the transition from a 1021 dentry (optional). This allows autofs, for example, to hold up clients 1022 waiting to explore behind a 'mountpoint' whilst letting the daemon go 1023 past and construct the subtree there. 0 should be returned to let the 1024 calling process continue. -EISDIR can be returned to tell pathwalk to 1025 use this directory as an ordinary directory and to ignore anything 1026 mounted on it and not to check the automount flag. Any other error 1027 code will abort pathwalk completely. 1028 1029 If the 'rcu_walk' parameter is true, then the caller is doing a 1030 pathwalk in RCU-walk mode. Sleeping is not permitted in this mode, 1031 and the caller can be asked to leave it and call again by returning 1032 -ECHILD. 1033 1034 This function is only used if DCACHE_MANAGE_TRANSIT is set on the 1035 dentry being transited from. 1036 1037Example : 1038 1039static char *pipefs_dname(struct dentry *dent, char *buffer, int buflen) 1040{ 1041 return dynamic_dname(dentry, buffer, buflen, "pipe:[%lu]", 1042 dentry->d_inode->i_ino); 1043} 1044 1045Each dentry has a pointer to its parent dentry, as well as a hash list 1046of child dentries. Child dentries are basically like files in a 1047directory. 1048 1049 1050Directory Entry Cache API 1051-------------------------- 1052 1053There are a number of functions defined which permit a filesystem to 1054manipulate dentries: 1055 1056 dget: open a new handle for an existing dentry (this just increments 1057 the usage count) 1058 1059 dput: close a handle for a dentry (decrements the usage count). If 1060 the usage count drops to 0, and the dentry is still in its 1061 parent's hash, the "d_delete" method is called to check whether 1062 it should be cached. If it should not be cached, or if the dentry 1063 is not hashed, it is deleted. Otherwise cached dentries are put 1064 into an LRU list to be reclaimed on memory shortage. 1065 1066 d_drop: this unhashes a dentry from its parents hash list. A 1067 subsequent call to dput() will deallocate the dentry if its 1068 usage count drops to 0 1069 1070 d_delete: delete a dentry. If there are no other open references to 1071 the dentry then the dentry is turned into a negative dentry 1072 (the d_iput() method is called). If there are other 1073 references, then d_drop() is called instead 1074 1075 d_add: add a dentry to its parents hash list and then calls 1076 d_instantiate() 1077 1078 d_instantiate: add a dentry to the alias hash list for the inode and 1079 updates the "d_inode" member. The "i_count" member in the 1080 inode structure should be set/incremented. If the inode 1081 pointer is NULL, the dentry is called a "negative 1082 dentry". This function is commonly called when an inode is 1083 created for an existing negative dentry 1084 1085 d_lookup: look up a dentry given its parent and path name component 1086 It looks up the child of that given name from the dcache 1087 hash table. If it is found, the reference count is incremented 1088 and the dentry is returned. The caller must use dput() 1089 to free the dentry when it finishes using it. 1090 1091Mount Options 1092============= 1093 1094Parsing options 1095--------------- 1096 1097On mount and remount the filesystem is passed a string containing a 1098comma separated list of mount options. The options can have either of 1099these forms: 1100 1101 option 1102 option=value 1103 1104The <linux/parser.h> header defines an API that helps parse these 1105options. There are plenty of examples on how to use it in existing 1106filesystems. 1107 1108Showing options 1109--------------- 1110 1111If a filesystem accepts mount options, it must define show_options() 1112to show all the currently active options. The rules are: 1113 1114 - options MUST be shown which are not default or their values differ 1115 from the default 1116 1117 - options MAY be shown which are enabled by default or have their 1118 default value 1119 1120Options used only internally between a mount helper and the kernel 1121(such as file descriptors), or which only have an effect during the 1122mounting (such as ones controlling the creation of a journal) are exempt 1123from the above rules. 1124 1125The underlying reason for the above rules is to make sure, that a 1126mount can be accurately replicated (e.g. umounting and mounting again) 1127based on the information found in /proc/mounts. 1128 1129A simple method of saving options at mount/remount time and showing 1130them is provided with the save_mount_options() and 1131generic_show_options() helper functions. Please note, that using 1132these may have drawbacks. For more info see header comments for these 1133functions in fs/namespace.c. 1134 1135Resources 1136========= 1137 1138(Note some of these resources are not up-to-date with the latest kernel 1139 version.) 1140 1141Creating Linux virtual filesystems. 2002 1142 <http://lwn.net/Articles/13325/> 1143 1144The Linux Virtual File-system Layer by Neil Brown. 1999 1145 <http://www.cse.unsw.edu.au/~neilb/oss/linux-commentary/vfs.html> 1146 1147A tour of the Linux VFS by Michael K. Johnson. 1996 1148 <http://www.tldp.org/LDP/khg/HyperNews/get/fs/vfstour.html> 1149 1150A small trail through the Linux kernel by Andries Brouwer. 2001 1151 <http://www.win.tue.nl/~aeb/linux/vfs/trail.html> 1152