1.. SPDX-License-Identifier: GPL-2.0 2 3========================================= 4Overview of the Linux Virtual File System 5========================================= 6 7Original author: Richard Gooch <rgooch@atnf.csiro.au> 8 9- Copyright (C) 1999 Richard Gooch 10- Copyright (C) 2005 Pekka Enberg 11 12 13Introduction 14============ 15 16The Virtual File System (also known as the Virtual Filesystem Switch) is 17the software layer in the kernel that provides the filesystem interface 18to userspace programs. It also provides an abstraction within the 19kernel which allows different filesystem implementations to coexist. 20 21VFS system calls open(2), stat(2), read(2), write(2), chmod(2) and so on 22are called from a process context. Filesystem locking is described in 23the document Documentation/filesystems/locking.rst. 24 25 26Directory Entry Cache (dcache) 27------------------------------ 28 29The VFS implements the open(2), stat(2), chmod(2), and similar system 30calls. The pathname argument that is passed to them is used by the VFS 31to search through the directory entry cache (also known as the dentry 32cache or dcache). This provides a very fast look-up mechanism to 33translate a pathname (filename) into a specific dentry. Dentries live 34in RAM and are never saved to disc: they exist only for performance. 35 36The dentry cache is meant to be a view into your entire filespace. As 37most computers cannot fit all dentries in the RAM at the same time, some 38bits of the cache are missing. In order to resolve your pathname into a 39dentry, the VFS may have to resort to creating dentries along the way, 40and then loading the inode. This is done by looking up the inode. 41 42 43The Inode Object 44---------------- 45 46An individual dentry usually has a pointer to an inode. Inodes are 47filesystem objects such as regular files, directories, FIFOs and other 48beasts. They live either on the disc (for block device filesystems) or 49in the memory (for pseudo filesystems). Inodes that live on the disc 50are copied into the memory when required and changes to the inode are 51written back to disc. A single inode can be pointed to by multiple 52dentries (hard links, for example, do this). 53 54To look up an inode requires that the VFS calls the lookup() method of 55the parent directory inode. This method is installed by the specific 56filesystem implementation that the inode lives in. Once the VFS has the 57required dentry (and hence the inode), we can do all those boring things 58like open(2) the file, or stat(2) it to peek at the inode data. The 59stat(2) operation is fairly simple: once the VFS has the dentry, it 60peeks at the inode data and passes some of it back to userspace. 61 62 63The File Object 64--------------- 65 66Opening a file requires another operation: allocation of a file 67structure (this is the kernel-side implementation of file descriptors). 68The freshly allocated file structure is initialized with a pointer to 69the dentry and a set of file operation member functions. These are 70taken from the inode data. The open() file method is then called so the 71specific filesystem implementation can do its work. You can see that 72this is another switch performed by the VFS. The file structure is 73placed into the file descriptor table for the process. 74 75Reading, writing and closing files (and other assorted VFS operations) 76is done by using the userspace file descriptor to grab the appropriate 77file structure, and then calling the required file structure method to 78do whatever is required. For as long as the file is open, it keeps the 79dentry in use, which in turn means that the VFS inode is still in use. 80 81 82Registering and Mounting a Filesystem 83===================================== 84 85To register and unregister a filesystem, use the following API 86functions: 87 88.. code-block:: c 89 90 #include <linux/fs.h> 91 92 extern int register_filesystem(struct file_system_type *); 93 extern int unregister_filesystem(struct file_system_type *); 94 95The passed struct file_system_type describes your filesystem. When a 96request is made to mount a filesystem onto a directory in your 97namespace, the VFS will call the appropriate mount() method for the 98specific filesystem. New vfsmount referring to the tree returned by 99->mount() will be attached to the mountpoint, so that when pathname 100resolution reaches the mountpoint it will jump into the root of that 101vfsmount. 102 103You can see all filesystems that are registered to the kernel in the 104file /proc/filesystems. 105 106 107struct file_system_type 108----------------------- 109 110This describes the filesystem. The following 111members are defined: 112 113.. code-block:: c 114 115 struct file_system_type { 116 const char *name; 117 int fs_flags; 118 int (*init_fs_context)(struct fs_context *); 119 const struct fs_parameter_spec *parameters; 120 struct dentry *(*mount) (struct file_system_type *, int, 121 const char *, void *); 122 void (*kill_sb) (struct super_block *); 123 struct module *owner; 124 struct file_system_type * next; 125 struct hlist_head fs_supers; 126 127 struct lock_class_key s_lock_key; 128 struct lock_class_key s_umount_key; 129 struct lock_class_key s_vfs_rename_key; 130 struct lock_class_key s_writers_key[SB_FREEZE_LEVELS]; 131 132 struct lock_class_key i_lock_key; 133 struct lock_class_key i_mutex_key; 134 struct lock_class_key invalidate_lock_key; 135 struct lock_class_key i_mutex_dir_key; 136 }; 137 138``name`` 139 the name of the filesystem type, such as "ext2", "iso9660", 140 "msdos" and so on 141 142``fs_flags`` 143 various flags (i.e. FS_REQUIRES_DEV, FS_NO_DCACHE, etc.) 144 145``init_fs_context`` 146 Initializes 'struct fs_context' ->ops and ->fs_private fields with 147 filesystem-specific data. 148 149``parameters`` 150 Pointer to the array of filesystem parameters descriptors 151 'struct fs_parameter_spec'. 152 More info in Documentation/filesystems/mount_api.rst. 153 154``mount`` 155 the method to call when a new instance of this filesystem should 156 be mounted 157 158``kill_sb`` 159 the method to call when an instance of this filesystem should be 160 shut down 161 162 163``owner`` 164 for internal VFS use: you should initialize this to THIS_MODULE 165 in most cases. 166 167``next`` 168 for internal VFS use: you should initialize this to NULL 169 170``fs_supers`` 171 for internal VFS use: hlist of filesystem instances (superblocks) 172 173 s_lock_key, s_umount_key, s_vfs_rename_key, s_writers_key, 174 i_lock_key, i_mutex_key, invalidate_lock_key, i_mutex_dir_key: lockdep-specific 175 176The mount() method has the following arguments: 177 178``struct file_system_type *fs_type`` 179 describes the filesystem, partly initialized by the specific 180 filesystem code 181 182``int flags`` 183 mount flags 184 185``const char *dev_name`` 186 the device name we are mounting. 187 188``void *data`` 189 arbitrary mount options, usually comes as an ASCII string (see 190 "Mount Options" section) 191 192The mount() method must return the root dentry of the tree requested by 193caller. An active reference to its superblock must be grabbed and the 194superblock must be locked. On failure it should return ERR_PTR(error). 195 196The arguments match those of mount(2) and their interpretation depends 197on filesystem type. E.g. for block filesystems, dev_name is interpreted 198as block device name, that device is opened and if it contains a 199suitable filesystem image the method creates and initializes struct 200super_block accordingly, returning its root dentry to caller. 201 202->mount() may choose to return a subtree of existing filesystem - it 203doesn't have to create a new one. The main result from the caller's 204point of view is a reference to dentry at the root of (sub)tree to be 205attached; creation of new superblock is a common side effect. 206 207The most interesting member of the superblock structure that the mount() 208method fills in is the "s_op" field. This is a pointer to a "struct 209super_operations" which describes the next level of the filesystem 210implementation. 211 212Usually, a filesystem uses one of the generic mount() implementations 213and provides a fill_super() callback instead. The generic variants are: 214 215``mount_bdev`` 216 mount a filesystem residing on a block device 217 218``mount_nodev`` 219 mount a filesystem that is not backed by a device 220 221``mount_single`` 222 mount a filesystem which shares the instance between all mounts 223 224A fill_super() callback implementation has the following arguments: 225 226``struct super_block *sb`` 227 the superblock structure. The callback must initialize this 228 properly. 229 230``void *data`` 231 arbitrary mount options, usually comes as an ASCII string (see 232 "Mount Options" section) 233 234``int silent`` 235 whether or not to be silent on error 236 237 238The Superblock Object 239===================== 240 241A superblock object represents a mounted filesystem. 242 243 244struct super_operations 245----------------------- 246 247This describes how the VFS can manipulate the superblock of your 248filesystem. As of kernel 2.6.22, the following members are defined: 249 250.. code-block:: c 251 252 struct super_operations { 253 struct inode *(*alloc_inode)(struct super_block *sb); 254 void (*destroy_inode)(struct inode *); 255 256 void (*dirty_inode) (struct inode *, int flags); 257 int (*write_inode) (struct inode *, int); 258 void (*drop_inode) (struct inode *); 259 void (*delete_inode) (struct inode *); 260 void (*put_super) (struct super_block *); 261 int (*sync_fs)(struct super_block *sb, int wait); 262 int (*freeze_fs) (struct super_block *); 263 int (*unfreeze_fs) (struct super_block *); 264 int (*statfs) (struct dentry *, struct kstatfs *); 265 int (*remount_fs) (struct super_block *, int *, char *); 266 void (*clear_inode) (struct inode *); 267 void (*umount_begin) (struct super_block *); 268 269 int (*show_options)(struct seq_file *, struct dentry *); 270 271 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t); 272 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t); 273 int (*nr_cached_objects)(struct super_block *); 274 void (*free_cached_objects)(struct super_block *, int); 275 }; 276 277All methods are called without any locks being held, unless otherwise 278noted. This means that most methods can block safely. All methods are 279only called from a process context (i.e. not from an interrupt handler 280or bottom half). 281 282``alloc_inode`` 283 this method is called by alloc_inode() to allocate memory for 284 struct inode and initialize it. If this function is not 285 defined, a simple 'struct inode' is allocated. Normally 286 alloc_inode will be used to allocate a larger structure which 287 contains a 'struct inode' embedded within it. 288 289``destroy_inode`` 290 this method is called by destroy_inode() to release resources 291 allocated for struct inode. It is only required if 292 ->alloc_inode was defined and simply undoes anything done by 293 ->alloc_inode. 294 295``dirty_inode`` 296 this method is called by the VFS when an inode is marked dirty. 297 This is specifically for the inode itself being marked dirty, 298 not its data. If the update needs to be persisted by fdatasync(), 299 then I_DIRTY_DATASYNC will be set in the flags argument. 300 I_DIRTY_TIME will be set in the flags in case lazytime is enabled 301 and struct inode has times updated since the last ->dirty_inode 302 call. 303 304``write_inode`` 305 this method is called when the VFS needs to write an inode to 306 disc. The second parameter indicates whether the write should 307 be synchronous or not, not all filesystems check this flag. 308 309``drop_inode`` 310 called when the last access to the inode is dropped, with the 311 inode->i_lock spinlock held. 312 313 This method should be either NULL (normal UNIX filesystem 314 semantics) or "generic_delete_inode" (for filesystems that do 315 not want to cache inodes - causing "delete_inode" to always be 316 called regardless of the value of i_nlink) 317 318 The "generic_delete_inode()" behavior is equivalent to the old 319 practice of using "force_delete" in the put_inode() case, but 320 does not have the races that the "force_delete()" approach had. 321 322``delete_inode`` 323 called when the VFS wants to delete an inode 324 325``put_super`` 326 called when the VFS wishes to free the superblock 327 (i.e. unmount). This is called with the superblock lock held 328 329``sync_fs`` 330 called when VFS is writing out all dirty data associated with a 331 superblock. The second parameter indicates whether the method 332 should wait until the write out has been completed. Optional. 333 334``freeze_fs`` 335 called when VFS is locking a filesystem and forcing it into a 336 consistent state. This method is currently used by the Logical 337 Volume Manager (LVM). 338 339``unfreeze_fs`` 340 called when VFS is unlocking a filesystem and making it writable 341 again. 342 343``statfs`` 344 called when the VFS needs to get filesystem statistics. 345 346``remount_fs`` 347 called when the filesystem is remounted. This is called with 348 the kernel lock held 349 350``clear_inode`` 351 called then the VFS clears the inode. Optional 352 353``umount_begin`` 354 called when the VFS is unmounting a filesystem. 355 356``show_options`` 357 called by the VFS to show mount options for /proc/<pid>/mounts. 358 (see "Mount Options" section) 359 360``quota_read`` 361 called by the VFS to read from filesystem quota file. 362 363``quota_write`` 364 called by the VFS to write to filesystem quota file. 365 366``nr_cached_objects`` 367 called by the sb cache shrinking function for the filesystem to 368 return the number of freeable cached objects it contains. 369 Optional. 370 371``free_cache_objects`` 372 called by the sb cache shrinking function for the filesystem to 373 scan the number of objects indicated to try to free them. 374 Optional, but any filesystem implementing this method needs to 375 also implement ->nr_cached_objects for it to be called 376 correctly. 377 378 We can't do anything with any errors that the filesystem might 379 encountered, hence the void return type. This will never be 380 called if the VM is trying to reclaim under GFP_NOFS conditions, 381 hence this method does not need to handle that situation itself. 382 383 Implementations must include conditional reschedule calls inside 384 any scanning loop that is done. This allows the VFS to 385 determine appropriate scan batch sizes without having to worry 386 about whether implementations will cause holdoff problems due to 387 large scan batch sizes. 388 389Whoever sets up the inode is responsible for filling in the "i_op" 390field. This is a pointer to a "struct inode_operations" which describes 391the methods that can be performed on individual inodes. 392 393 394struct xattr_handlers 395--------------------- 396 397On filesystems that support extended attributes (xattrs), the s_xattr 398superblock field points to a NULL-terminated array of xattr handlers. 399Extended attributes are name:value pairs. 400 401``name`` 402 Indicates that the handler matches attributes with the specified 403 name (such as "system.posix_acl_access"); the prefix field must 404 be NULL. 405 406``prefix`` 407 Indicates that the handler matches all attributes with the 408 specified name prefix (such as "user."); the name field must be 409 NULL. 410 411``list`` 412 Determine if attributes matching this xattr handler should be 413 listed for a particular dentry. Used by some listxattr 414 implementations like generic_listxattr. 415 416``get`` 417 Called by the VFS to get the value of a particular extended 418 attribute. This method is called by the getxattr(2) system 419 call. 420 421``set`` 422 Called by the VFS to set the value of a particular extended 423 attribute. When the new value is NULL, called to remove a 424 particular extended attribute. This method is called by the 425 setxattr(2) and removexattr(2) system calls. 426 427When none of the xattr handlers of a filesystem match the specified 428attribute name or when a filesystem doesn't support extended attributes, 429the various ``*xattr(2)`` system calls return -EOPNOTSUPP. 430 431 432The Inode Object 433================ 434 435An inode object represents an object within the filesystem. 436 437 438struct inode_operations 439----------------------- 440 441This describes how the VFS can manipulate an inode in your filesystem. 442As of kernel 2.6.22, the following members are defined: 443 444.. code-block:: c 445 446 struct inode_operations { 447 int (*create) (struct mnt_idmap *, struct inode *,struct dentry *, umode_t, bool); 448 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int); 449 int (*link) (struct dentry *,struct inode *,struct dentry *); 450 int (*unlink) (struct inode *,struct dentry *); 451 int (*symlink) (struct mnt_idmap *, struct inode *,struct dentry *,const char *); 452 int (*mkdir) (struct mnt_idmap *, struct inode *,struct dentry *,umode_t); 453 int (*rmdir) (struct inode *,struct dentry *); 454 int (*mknod) (struct mnt_idmap *, struct inode *,struct dentry *,umode_t,dev_t); 455 int (*rename) (struct mnt_idmap *, struct inode *, struct dentry *, 456 struct inode *, struct dentry *, unsigned int); 457 int (*readlink) (struct dentry *, char __user *,int); 458 const char *(*get_link) (struct dentry *, struct inode *, 459 struct delayed_call *); 460 int (*permission) (struct mnt_idmap *, struct inode *, int); 461 struct posix_acl * (*get_inode_acl)(struct inode *, int, bool); 462 int (*setattr) (struct mnt_idmap *, struct dentry *, struct iattr *); 463 int (*getattr) (struct mnt_idmap *, const struct path *, struct kstat *, u32, unsigned int); 464 ssize_t (*listxattr) (struct dentry *, char *, size_t); 465 void (*update_time)(struct inode *, struct timespec *, int); 466 int (*atomic_open)(struct inode *, struct dentry *, struct file *, 467 unsigned open_flag, umode_t create_mode); 468 int (*tmpfile) (struct mnt_idmap *, struct inode *, struct file *, umode_t); 469 struct posix_acl * (*get_acl)(struct mnt_idmap *, struct dentry *, int); 470 int (*set_acl)(struct mnt_idmap *, struct dentry *, struct posix_acl *, int); 471 int (*fileattr_set)(struct mnt_idmap *idmap, 472 struct dentry *dentry, struct fileattr *fa); 473 int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa); 474 }; 475 476Again, all methods are called without any locks being held, unless 477otherwise noted. 478 479``create`` 480 called by the open(2) and creat(2) system calls. Only required 481 if you want to support regular files. The dentry you get should 482 not have an inode (i.e. it should be a negative dentry). Here 483 you will probably call d_instantiate() with the dentry and the 484 newly created inode 485 486``lookup`` 487 called when the VFS needs to look up an inode in a parent 488 directory. The name to look for is found in the dentry. This 489 method must call d_add() to insert the found inode into the 490 dentry. The "i_count" field in the inode structure should be 491 incremented. If the named inode does not exist a NULL inode 492 should be inserted into the dentry (this is called a negative 493 dentry). Returning an error code from this routine must only be 494 done on a real error, otherwise creating inodes with system 495 calls like create(2), mknod(2), mkdir(2) and so on will fail. 496 If you wish to overload the dentry methods then you should 497 initialise the "d_dop" field in the dentry; this is a pointer to 498 a struct "dentry_operations". This method is called with the 499 directory inode semaphore held 500 501``link`` 502 called by the link(2) system call. Only required if you want to 503 support hard links. You will probably need to call 504 d_instantiate() just as you would in the create() method 505 506``unlink`` 507 called by the unlink(2) system call. Only required if you want 508 to support deleting inodes 509 510``symlink`` 511 called by the symlink(2) system call. Only required if you want 512 to support symlinks. You will probably need to call 513 d_instantiate() just as you would in the create() method 514 515``mkdir`` 516 called by the mkdir(2) system call. Only required if you want 517 to support creating subdirectories. You will probably need to 518 call d_instantiate() just as you would in the create() method 519 520``rmdir`` 521 called by the rmdir(2) system call. Only required if you want 522 to support deleting subdirectories 523 524``mknod`` 525 called by the mknod(2) system call to create a device (char, 526 block) inode or a named pipe (FIFO) or socket. Only required if 527 you want to support creating these types of inodes. You will 528 probably need to call d_instantiate() just as you would in the 529 create() method 530 531``rename`` 532 called by the rename(2) system call to rename the object to have 533 the parent and name given by the second inode and dentry. 534 535 The filesystem must return -EINVAL for any unsupported or 536 unknown flags. Currently the following flags are implemented: 537 (1) RENAME_NOREPLACE: this flag indicates that if the target of 538 the rename exists the rename should fail with -EEXIST instead of 539 replacing the target. The VFS already checks for existence, so 540 for local filesystems the RENAME_NOREPLACE implementation is 541 equivalent to plain rename. 542 (2) RENAME_EXCHANGE: exchange source and target. Both must 543 exist; this is checked by the VFS. Unlike plain rename, source 544 and target may be of different type. 545 546``get_link`` 547 called by the VFS to follow a symbolic link to the inode it 548 points to. Only required if you want to support symbolic links. 549 This method returns the symlink body to traverse (and possibly 550 resets the current position with nd_jump_link()). If the body 551 won't go away until the inode is gone, nothing else is needed; 552 if it needs to be otherwise pinned, arrange for its release by 553 having get_link(..., ..., done) do set_delayed_call(done, 554 destructor, argument). In that case destructor(argument) will 555 be called once VFS is done with the body you've returned. May 556 be called in RCU mode; that is indicated by NULL dentry 557 argument. If request can't be handled without leaving RCU mode, 558 have it return ERR_PTR(-ECHILD). 559 560 If the filesystem stores the symlink target in ->i_link, the 561 VFS may use it directly without calling ->get_link(); however, 562 ->get_link() must still be provided. ->i_link must not be 563 freed until after an RCU grace period. Writing to ->i_link 564 post-iget() time requires a 'release' memory barrier. 565 566``readlink`` 567 this is now just an override for use by readlink(2) for the 568 cases when ->get_link uses nd_jump_link() or object is not in 569 fact a symlink. Normally filesystems should only implement 570 ->get_link for symlinks and readlink(2) will automatically use 571 that. 572 573``permission`` 574 called by the VFS to check for access rights on a POSIX-like 575 filesystem. 576 577 May be called in rcu-walk mode (mask & MAY_NOT_BLOCK). If in 578 rcu-walk mode, the filesystem must check the permission without 579 blocking or storing to the inode. 580 581 If a situation is encountered that rcu-walk cannot handle, 582 return 583 -ECHILD and it will be called again in ref-walk mode. 584 585``setattr`` 586 called by the VFS to set attributes for a file. This method is 587 called by chmod(2) and related system calls. 588 589``getattr`` 590 called by the VFS to get attributes of a file. This method is 591 called by stat(2) and related system calls. 592 593``listxattr`` 594 called by the VFS to list all extended attributes for a given 595 file. This method is called by the listxattr(2) system call. 596 597``update_time`` 598 called by the VFS to update a specific time or the i_version of 599 an inode. If this is not defined the VFS will update the inode 600 itself and call mark_inode_dirty_sync. 601 602``atomic_open`` 603 called on the last component of an open. Using this optional 604 method the filesystem can look up, possibly create and open the 605 file in one atomic operation. If it wants to leave actual 606 opening to the caller (e.g. if the file turned out to be a 607 symlink, device, or just something filesystem won't do atomic 608 open for), it may signal this by returning finish_no_open(file, 609 dentry). This method is only called if the last component is 610 negative or needs lookup. Cached positive dentries are still 611 handled by f_op->open(). If the file was created, FMODE_CREATED 612 flag should be set in file->f_mode. In case of O_EXCL the 613 method must only succeed if the file didn't exist and hence 614 FMODE_CREATED shall always be set on success. 615 616``tmpfile`` 617 called in the end of O_TMPFILE open(). Optional, equivalent to 618 atomically creating, opening and unlinking a file in given 619 directory. On success needs to return with the file already 620 open; this can be done by calling finish_open_simple() right at 621 the end. 622 623``fileattr_get`` 624 called on ioctl(FS_IOC_GETFLAGS) and ioctl(FS_IOC_FSGETXATTR) to 625 retrieve miscellaneous file flags and attributes. Also called 626 before the relevant SET operation to check what is being changed 627 (in this case with i_rwsem locked exclusive). If unset, then 628 fall back to f_op->ioctl(). 629 630``fileattr_set`` 631 called on ioctl(FS_IOC_SETFLAGS) and ioctl(FS_IOC_FSSETXATTR) to 632 change miscellaneous file flags and attributes. Callers hold 633 i_rwsem exclusive. If unset, then fall back to f_op->ioctl(). 634 635 636The Address Space Object 637======================== 638 639The address space object is used to group and manage pages in the page 640cache. It can be used to keep track of the pages in a file (or anything 641else) and also track the mapping of sections of the file into process 642address spaces. 643 644There are a number of distinct yet related services that an 645address-space can provide. These include communicating memory pressure, 646page lookup by address, and keeping track of pages tagged as Dirty or 647Writeback. 648 649The first can be used independently to the others. The VM can try to 650either write dirty pages in order to clean them, or release clean pages 651in order to reuse them. To do this it can call the ->writepage method 652on dirty pages, and ->release_folio on clean folios with the private 653flag set. Clean pages without PagePrivate and with no external references 654will be released without notice being given to the address_space. 655 656To achieve this functionality, pages need to be placed on an LRU with 657lru_cache_add and mark_page_active needs to be called whenever the page 658is used. 659 660Pages are normally kept in a radix tree index by ->index. This tree 661maintains information about the PG_Dirty and PG_Writeback status of each 662page, so that pages with either of these flags can be found quickly. 663 664The Dirty tag is primarily used by mpage_writepages - the default 665->writepages method. It uses the tag to find dirty pages to call 666->writepage on. If mpage_writepages is not used (i.e. the address 667provides its own ->writepages) , the PAGECACHE_TAG_DIRTY tag is almost 668unused. write_inode_now and sync_inode do use it (through 669__sync_single_inode) to check if ->writepages has been successful in 670writing out the whole address_space. 671 672The Writeback tag is used by filemap*wait* and sync_page* functions, via 673filemap_fdatawait_range, to wait for all writeback to complete. 674 675An address_space handler may attach extra information to a page, 676typically using the 'private' field in the 'struct page'. If such 677information is attached, the PG_Private flag should be set. This will 678cause various VM routines to make extra calls into the address_space 679handler to deal with that data. 680 681An address space acts as an intermediate between storage and 682application. Data is read into the address space a whole page at a 683time, and provided to the application either by copying of the page, or 684by memory-mapping the page. Data is written into the address space by 685the application, and then written-back to storage typically in whole 686pages, however the address_space has finer control of write sizes. 687 688The read process essentially only requires 'read_folio'. The write 689process is more complicated and uses write_begin/write_end or 690dirty_folio to write data into the address_space, and writepage and 691writepages to writeback data to storage. 692 693Adding and removing pages to/from an address_space is protected by the 694inode's i_mutex. 695 696When data is written to a page, the PG_Dirty flag should be set. It 697typically remains set until writepage asks for it to be written. This 698should clear PG_Dirty and set PG_Writeback. It can be actually written 699at any point after PG_Dirty is clear. Once it is known to be safe, 700PG_Writeback is cleared. 701 702Writeback makes use of a writeback_control structure to direct the 703operations. This gives the writepage and writepages operations some 704information about the nature of and reason for the writeback request, 705and the constraints under which it is being done. It is also used to 706return information back to the caller about the result of a writepage or 707writepages request. 708 709 710Handling errors during writeback 711-------------------------------- 712 713Most applications that do buffered I/O will periodically call a file 714synchronization call (fsync, fdatasync, msync or sync_file_range) to 715ensure that data written has made it to the backing store. When there 716is an error during writeback, they expect that error to be reported when 717a file sync request is made. After an error has been reported on one 718request, subsequent requests on the same file descriptor should return 7190, unless further writeback errors have occurred since the previous file 720syncronization. 721 722Ideally, the kernel would report errors only on file descriptions on 723which writes were done that subsequently failed to be written back. The 724generic pagecache infrastructure does not track the file descriptions 725that have dirtied each individual page however, so determining which 726file descriptors should get back an error is not possible. 727 728Instead, the generic writeback error tracking infrastructure in the 729kernel settles for reporting errors to fsync on all file descriptions 730that were open at the time that the error occurred. In a situation with 731multiple writers, all of them will get back an error on a subsequent 732fsync, even if all of the writes done through that particular file 733descriptor succeeded (or even if there were no writes on that file 734descriptor at all). 735 736Filesystems that wish to use this infrastructure should call 737mapping_set_error to record the error in the address_space when it 738occurs. Then, after writing back data from the pagecache in their 739file->fsync operation, they should call file_check_and_advance_wb_err to 740ensure that the struct file's error cursor has advanced to the correct 741point in the stream of errors emitted by the backing device(s). 742 743 744struct address_space_operations 745------------------------------- 746 747This describes how the VFS can manipulate mapping of a file to page 748cache in your filesystem. The following members are defined: 749 750.. code-block:: c 751 752 struct address_space_operations { 753 int (*writepage)(struct page *page, struct writeback_control *wbc); 754 int (*read_folio)(struct file *, struct folio *); 755 int (*writepages)(struct address_space *, struct writeback_control *); 756 bool (*dirty_folio)(struct address_space *, struct folio *); 757 void (*readahead)(struct readahead_control *); 758 int (*write_begin)(struct file *, struct address_space *mapping, 759 loff_t pos, unsigned len, 760 struct page **pagep, void **fsdata); 761 int (*write_end)(struct file *, struct address_space *mapping, 762 loff_t pos, unsigned len, unsigned copied, 763 struct page *page, void *fsdata); 764 sector_t (*bmap)(struct address_space *, sector_t); 765 void (*invalidate_folio) (struct folio *, size_t start, size_t len); 766 bool (*release_folio)(struct folio *, gfp_t); 767 void (*free_folio)(struct folio *); 768 ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter); 769 int (*migrate_folio)(struct mapping *, struct folio *dst, 770 struct folio *src, enum migrate_mode); 771 int (*launder_folio) (struct folio *); 772 773 bool (*is_partially_uptodate) (struct folio *, size_t from, 774 size_t count); 775 void (*is_dirty_writeback)(struct folio *, bool *, bool *); 776 int (*error_remove_page) (struct mapping *mapping, struct page *page); 777 int (*swap_activate)(struct swap_info_struct *sis, struct file *f, sector_t *span) 778 int (*swap_deactivate)(struct file *); 779 int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter); 780 }; 781 782``writepage`` 783 called by the VM to write a dirty page to backing store. This 784 may happen for data integrity reasons (i.e. 'sync'), or to free 785 up memory (flush). The difference can be seen in 786 wbc->sync_mode. The PG_Dirty flag has been cleared and 787 PageLocked is true. writepage should start writeout, should set 788 PG_Writeback, and should make sure the page is unlocked, either 789 synchronously or asynchronously when the write operation 790 completes. 791 792 If wbc->sync_mode is WB_SYNC_NONE, ->writepage doesn't have to 793 try too hard if there are problems, and may choose to write out 794 other pages from the mapping if that is easier (e.g. due to 795 internal dependencies). If it chooses not to start writeout, it 796 should return AOP_WRITEPAGE_ACTIVATE so that the VM will not 797 keep calling ->writepage on that page. 798 799 See the file "Locking" for more details. 800 801``read_folio`` 802 Called by the page cache to read a folio from the backing store. 803 The 'file' argument supplies authentication information to network 804 filesystems, and is generally not used by block based filesystems. 805 It may be NULL if the caller does not have an open file (eg if 806 the kernel is performing a read for itself rather than on behalf 807 of a userspace process with an open file). 808 809 If the mapping does not support large folios, the folio will 810 contain a single page. The folio will be locked when read_folio 811 is called. If the read completes successfully, the folio should 812 be marked uptodate. The filesystem should unlock the folio 813 once the read has completed, whether it was successful or not. 814 The filesystem does not need to modify the refcount on the folio; 815 the page cache holds a reference count and that will not be 816 released until the folio is unlocked. 817 818 Filesystems may implement ->read_folio() synchronously. 819 In normal operation, folios are read through the ->readahead() 820 method. Only if this fails, or if the caller needs to wait for 821 the read to complete will the page cache call ->read_folio(). 822 Filesystems should not attempt to perform their own readahead 823 in the ->read_folio() operation. 824 825 If the filesystem cannot perform the read at this time, it can 826 unlock the folio, do whatever action it needs to ensure that the 827 read will succeed in the future and return AOP_TRUNCATED_PAGE. 828 In this case, the caller should look up the folio, lock it, 829 and call ->read_folio again. 830 831 Callers may invoke the ->read_folio() method directly, but using 832 read_mapping_folio() will take care of locking, waiting for the 833 read to complete and handle cases such as AOP_TRUNCATED_PAGE. 834 835``writepages`` 836 called by the VM to write out pages associated with the 837 address_space object. If wbc->sync_mode is WB_SYNC_ALL, then 838 the writeback_control will specify a range of pages that must be 839 written out. If it is WB_SYNC_NONE, then a nr_to_write is 840 given and that many pages should be written if possible. If no 841 ->writepages is given, then mpage_writepages is used instead. 842 This will choose pages from the address space that are tagged as 843 DIRTY and will pass them to ->writepage. 844 845``dirty_folio`` 846 called by the VM to mark a folio as dirty. This is particularly 847 needed if an address space attaches private data to a folio, and 848 that data needs to be updated when a folio is dirtied. This is 849 called, for example, when a memory mapped page gets modified. 850 If defined, it should set the folio dirty flag, and the 851 PAGECACHE_TAG_DIRTY search mark in i_pages. 852 853``readahead`` 854 Called by the VM to read pages associated with the address_space 855 object. The pages are consecutive in the page cache and are 856 locked. The implementation should decrement the page refcount 857 after starting I/O on each page. Usually the page will be 858 unlocked by the I/O completion handler. The set of pages are 859 divided into some sync pages followed by some async pages, 860 rac->ra->async_size gives the number of async pages. The 861 filesystem should attempt to read all sync pages but may decide 862 to stop once it reaches the async pages. If it does decide to 863 stop attempting I/O, it can simply return. The caller will 864 remove the remaining pages from the address space, unlock them 865 and decrement the page refcount. Set PageUptodate if the I/O 866 completes successfully. Setting PageError on any page will be 867 ignored; simply unlock the page if an I/O error occurs. 868 869``write_begin`` 870 Called by the generic buffered write code to ask the filesystem 871 to prepare to write len bytes at the given offset in the file. 872 The address_space should check that the write will be able to 873 complete, by allocating space if necessary and doing any other 874 internal housekeeping. If the write will update parts of any 875 basic-blocks on storage, then those blocks should be pre-read 876 (if they haven't been read already) so that the updated blocks 877 can be written out properly. 878 879 The filesystem must return the locked pagecache page for the 880 specified offset, in ``*pagep``, for the caller to write into. 881 882 It must be able to cope with short writes (where the length 883 passed to write_begin is greater than the number of bytes copied 884 into the page). 885 886 A void * may be returned in fsdata, which then gets passed into 887 write_end. 888 889 Returns 0 on success; < 0 on failure (which is the error code), 890 in which case write_end is not called. 891 892``write_end`` 893 After a successful write_begin, and data copy, write_end must be 894 called. len is the original len passed to write_begin, and 895 copied is the amount that was able to be copied. 896 897 The filesystem must take care of unlocking the page and 898 releasing it refcount, and updating i_size. 899 900 Returns < 0 on failure, otherwise the number of bytes (<= 901 'copied') that were able to be copied into pagecache. 902 903``bmap`` 904 called by the VFS to map a logical block offset within object to 905 physical block number. This method is used by the FIBMAP ioctl 906 and for working with swap-files. To be able to swap to a file, 907 the file must have a stable mapping to a block device. The swap 908 system does not go through the filesystem but instead uses bmap 909 to find out where the blocks in the file are and uses those 910 addresses directly. 911 912``invalidate_folio`` 913 If a folio has private data, then invalidate_folio will be 914 called when part or all of the folio is to be removed from the 915 address space. This generally corresponds to either a 916 truncation, punch hole or a complete invalidation of the address 917 space (in the latter case 'offset' will always be 0 and 'length' 918 will be folio_size()). Any private data associated with the folio 919 should be updated to reflect this truncation. If offset is 0 920 and length is folio_size(), then the private data should be 921 released, because the folio must be able to be completely 922 discarded. This may be done by calling the ->release_folio 923 function, but in this case the release MUST succeed. 924 925``release_folio`` 926 release_folio is called on folios with private data to tell the 927 filesystem that the folio is about to be freed. ->release_folio 928 should remove any private data from the folio and clear the 929 private flag. If release_folio() fails, it should return false. 930 release_folio() is used in two distinct though related cases. 931 The first is when the VM wants to free a clean folio with no 932 active users. If ->release_folio succeeds, the folio will be 933 removed from the address_space and be freed. 934 935 The second case is when a request has been made to invalidate 936 some or all folios in an address_space. This can happen 937 through the fadvise(POSIX_FADV_DONTNEED) system call or by the 938 filesystem explicitly requesting it as nfs and 9p do (when they 939 believe the cache may be out of date with storage) by calling 940 invalidate_inode_pages2(). If the filesystem makes such a call, 941 and needs to be certain that all folios are invalidated, then 942 its release_folio will need to ensure this. Possibly it can 943 clear the uptodate flag if it cannot free private data yet. 944 945``free_folio`` 946 free_folio is called once the folio is no longer visible in the 947 page cache in order to allow the cleanup of any private data. 948 Since it may be called by the memory reclaimer, it should not 949 assume that the original address_space mapping still exists, and 950 it should not block. 951 952``direct_IO`` 953 called by the generic read/write routines to perform direct_IO - 954 that is IO requests which bypass the page cache and transfer 955 data directly between the storage and the application's address 956 space. 957 958``migrate_folio`` 959 This is used to compact the physical memory usage. If the VM 960 wants to relocate a folio (maybe from a memory device that is 961 signalling imminent failure) it will pass a new folio and an old 962 folio to this function. migrate_folio should transfer any private 963 data across and update any references that it has to the folio. 964 965``launder_folio`` 966 Called before freeing a folio - it writes back the dirty folio. 967 To prevent redirtying the folio, it is kept locked during the 968 whole operation. 969 970``is_partially_uptodate`` 971 Called by the VM when reading a file through the pagecache when 972 the underlying blocksize is smaller than the size of the folio. 973 If the required block is up to date then the read can complete 974 without needing I/O to bring the whole page up to date. 975 976``is_dirty_writeback`` 977 Called by the VM when attempting to reclaim a folio. The VM uses 978 dirty and writeback information to determine if it needs to 979 stall to allow flushers a chance to complete some IO. 980 Ordinarily it can use folio_test_dirty and folio_test_writeback but 981 some filesystems have more complex state (unstable folios in NFS 982 prevent reclaim) or do not set those flags due to locking 983 problems. This callback allows a filesystem to indicate to the 984 VM if a folio should be treated as dirty or writeback for the 985 purposes of stalling. 986 987``error_remove_page`` 988 normally set to generic_error_remove_page if truncation is ok 989 for this address space. Used for memory failure handling. 990 Setting this implies you deal with pages going away under you, 991 unless you have them locked or reference counts increased. 992 993``swap_activate`` 994 995 Called to prepare the given file for swap. It should perform 996 any validation and preparation necessary to ensure that writes 997 can be performed with minimal memory allocation. It should call 998 add_swap_extent(), or the helper iomap_swapfile_activate(), and 999 return the number of extents added. If IO should be submitted 1000 through ->swap_rw(), it should set SWP_FS_OPS, otherwise IO will 1001 be submitted directly to the block device ``sis->bdev``. 1002 1003``swap_deactivate`` 1004 Called during swapoff on files where swap_activate was 1005 successful. 1006 1007``swap_rw`` 1008 Called to read or write swap pages when SWP_FS_OPS is set. 1009 1010The File Object 1011=============== 1012 1013A file object represents a file opened by a process. This is also known 1014as an "open file description" in POSIX parlance. 1015 1016 1017struct file_operations 1018---------------------- 1019 1020This describes how the VFS can manipulate an open file. As of kernel 10214.18, the following members are defined: 1022 1023.. code-block:: c 1024 1025 struct file_operations { 1026 struct module *owner; 1027 loff_t (*llseek) (struct file *, loff_t, int); 1028 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *); 1029 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *); 1030 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *); 1031 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *); 1032 int (*iopoll)(struct kiocb *kiocb, bool spin); 1033 int (*iterate) (struct file *, struct dir_context *); 1034 int (*iterate_shared) (struct file *, struct dir_context *); 1035 __poll_t (*poll) (struct file *, struct poll_table_struct *); 1036 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long); 1037 long (*compat_ioctl) (struct file *, unsigned int, unsigned long); 1038 int (*mmap) (struct file *, struct vm_area_struct *); 1039 int (*open) (struct inode *, struct file *); 1040 int (*flush) (struct file *, fl_owner_t id); 1041 int (*release) (struct inode *, struct file *); 1042 int (*fsync) (struct file *, loff_t, loff_t, int datasync); 1043 int (*fasync) (int, struct file *, int); 1044 int (*lock) (struct file *, int, struct file_lock *); 1045 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int); 1046 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 1047 int (*check_flags)(int); 1048 int (*flock) (struct file *, int, struct file_lock *); 1049 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int); 1050 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int); 1051 int (*setlease)(struct file *, long, struct file_lock **, void **); 1052 long (*fallocate)(struct file *file, int mode, loff_t offset, 1053 loff_t len); 1054 void (*show_fdinfo)(struct seq_file *m, struct file *f); 1055 #ifndef CONFIG_MMU 1056 unsigned (*mmap_capabilities)(struct file *); 1057 #endif 1058 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *, loff_t, size_t, unsigned int); 1059 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in, 1060 struct file *file_out, loff_t pos_out, 1061 loff_t len, unsigned int remap_flags); 1062 int (*fadvise)(struct file *, loff_t, loff_t, int); 1063 }; 1064 1065Again, all methods are called without any locks being held, unless 1066otherwise noted. 1067 1068``llseek`` 1069 called when the VFS needs to move the file position index 1070 1071``read`` 1072 called by read(2) and related system calls 1073 1074``read_iter`` 1075 possibly asynchronous read with iov_iter as destination 1076 1077``write`` 1078 called by write(2) and related system calls 1079 1080``write_iter`` 1081 possibly asynchronous write with iov_iter as source 1082 1083``iopoll`` 1084 called when aio wants to poll for completions on HIPRI iocbs 1085 1086``iterate`` 1087 called when the VFS needs to read the directory contents 1088 1089``iterate_shared`` 1090 called when the VFS needs to read the directory contents when 1091 filesystem supports concurrent dir iterators 1092 1093``poll`` 1094 called by the VFS when a process wants to check if there is 1095 activity on this file and (optionally) go to sleep until there 1096 is activity. Called by the select(2) and poll(2) system calls 1097 1098``unlocked_ioctl`` 1099 called by the ioctl(2) system call. 1100 1101``compat_ioctl`` 1102 called by the ioctl(2) system call when 32 bit system calls are 1103 used on 64 bit kernels. 1104 1105``mmap`` 1106 called by the mmap(2) system call 1107 1108``open`` 1109 called by the VFS when an inode should be opened. When the VFS 1110 opens a file, it creates a new "struct file". It then calls the 1111 open method for the newly allocated file structure. You might 1112 think that the open method really belongs in "struct 1113 inode_operations", and you may be right. I think it's done the 1114 way it is because it makes filesystems simpler to implement. 1115 The open() method is a good place to initialize the 1116 "private_data" member in the file structure if you want to point 1117 to a device structure 1118 1119``flush`` 1120 called by the close(2) system call to flush a file 1121 1122``release`` 1123 called when the last reference to an open file is closed 1124 1125``fsync`` 1126 called by the fsync(2) system call. Also see the section above 1127 entitled "Handling errors during writeback". 1128 1129``fasync`` 1130 called by the fcntl(2) system call when asynchronous 1131 (non-blocking) mode is enabled for a file 1132 1133``lock`` 1134 called by the fcntl(2) system call for F_GETLK, F_SETLK, and 1135 F_SETLKW commands 1136 1137``get_unmapped_area`` 1138 called by the mmap(2) system call 1139 1140``check_flags`` 1141 called by the fcntl(2) system call for F_SETFL command 1142 1143``flock`` 1144 called by the flock(2) system call 1145 1146``splice_write`` 1147 called by the VFS to splice data from a pipe to a file. This 1148 method is used by the splice(2) system call 1149 1150``splice_read`` 1151 called by the VFS to splice data from file to a pipe. This 1152 method is used by the splice(2) system call 1153 1154``setlease`` 1155 called by the VFS to set or release a file lock lease. setlease 1156 implementations should call generic_setlease to record or remove 1157 the lease in the inode after setting it. 1158 1159``fallocate`` 1160 called by the VFS to preallocate blocks or punch a hole. 1161 1162``copy_file_range`` 1163 called by the copy_file_range(2) system call. 1164 1165``remap_file_range`` 1166 called by the ioctl(2) system call for FICLONERANGE and FICLONE 1167 and FIDEDUPERANGE commands to remap file ranges. An 1168 implementation should remap len bytes at pos_in of the source 1169 file into the dest file at pos_out. Implementations must handle 1170 callers passing in len == 0; this means "remap to the end of the 1171 source file". The return value should the number of bytes 1172 remapped, or the usual negative error code if errors occurred 1173 before any bytes were remapped. The remap_flags parameter 1174 accepts REMAP_FILE_* flags. If REMAP_FILE_DEDUP is set then the 1175 implementation must only remap if the requested file ranges have 1176 identical contents. If REMAP_FILE_CAN_SHORTEN is set, the caller is 1177 ok with the implementation shortening the request length to 1178 satisfy alignment or EOF requirements (or any other reason). 1179 1180``fadvise`` 1181 possibly called by the fadvise64() system call. 1182 1183Note that the file operations are implemented by the specific 1184filesystem in which the inode resides. When opening a device node 1185(character or block special) most filesystems will call special 1186support routines in the VFS which will locate the required device 1187driver information. These support routines replace the filesystem file 1188operations with those for the device driver, and then proceed to call 1189the new open() method for the file. This is how opening a device file 1190in the filesystem eventually ends up calling the device driver open() 1191method. 1192 1193 1194Directory Entry Cache (dcache) 1195============================== 1196 1197 1198struct dentry_operations 1199------------------------ 1200 1201This describes how a filesystem can overload the standard dentry 1202operations. Dentries and the dcache are the domain of the VFS and the 1203individual filesystem implementations. Device drivers have no business 1204here. These methods may be set to NULL, as they are either optional or 1205the VFS uses a default. As of kernel 2.6.22, the following members are 1206defined: 1207 1208.. code-block:: c 1209 1210 struct dentry_operations { 1211 int (*d_revalidate)(struct dentry *, unsigned int); 1212 int (*d_weak_revalidate)(struct dentry *, unsigned int); 1213 int (*d_hash)(const struct dentry *, struct qstr *); 1214 int (*d_compare)(const struct dentry *, 1215 unsigned int, const char *, const struct qstr *); 1216 int (*d_delete)(const struct dentry *); 1217 int (*d_init)(struct dentry *); 1218 void (*d_release)(struct dentry *); 1219 void (*d_iput)(struct dentry *, struct inode *); 1220 char *(*d_dname)(struct dentry *, char *, int); 1221 struct vfsmount *(*d_automount)(struct path *); 1222 int (*d_manage)(const struct path *, bool); 1223 struct dentry *(*d_real)(struct dentry *, const struct inode *); 1224 }; 1225 1226``d_revalidate`` 1227 called when the VFS needs to revalidate a dentry. This is 1228 called whenever a name look-up finds a dentry in the dcache. 1229 Most local filesystems leave this as NULL, because all their 1230 dentries in the dcache are valid. Network filesystems are 1231 different since things can change on the server without the 1232 client necessarily being aware of it. 1233 1234 This function should return a positive value if the dentry is 1235 still valid, and zero or a negative error code if it isn't. 1236 1237 d_revalidate may be called in rcu-walk mode (flags & 1238 LOOKUP_RCU). If in rcu-walk mode, the filesystem must 1239 revalidate the dentry without blocking or storing to the dentry, 1240 d_parent and d_inode should not be used without care (because 1241 they can change and, in d_inode case, even become NULL under 1242 us). 1243 1244 If a situation is encountered that rcu-walk cannot handle, 1245 return 1246 -ECHILD and it will be called again in ref-walk mode. 1247 1248``d_weak_revalidate`` 1249 called when the VFS needs to revalidate a "jumped" dentry. This 1250 is called when a path-walk ends at dentry that was not acquired 1251 by doing a lookup in the parent directory. This includes "/", 1252 "." and "..", as well as procfs-style symlinks and mountpoint 1253 traversal. 1254 1255 In this case, we are less concerned with whether the dentry is 1256 still fully correct, but rather that the inode is still valid. 1257 As with d_revalidate, most local filesystems will set this to 1258 NULL since their dcache entries are always valid. 1259 1260 This function has the same return code semantics as 1261 d_revalidate. 1262 1263 d_weak_revalidate is only called after leaving rcu-walk mode. 1264 1265``d_hash`` 1266 called when the VFS adds a dentry to the hash table. The first 1267 dentry passed to d_hash is the parent directory that the name is 1268 to be hashed into. 1269 1270 Same locking and synchronisation rules as d_compare regarding 1271 what is safe to dereference etc. 1272 1273``d_compare`` 1274 called to compare a dentry name with a given name. The first 1275 dentry is the parent of the dentry to be compared, the second is 1276 the child dentry. len and name string are properties of the 1277 dentry to be compared. qstr is the name to compare it with. 1278 1279 Must be constant and idempotent, and should not take locks if 1280 possible, and should not or store into the dentry. Should not 1281 dereference pointers outside the dentry without lots of care 1282 (eg. d_parent, d_inode, d_name should not be used). 1283 1284 However, our vfsmount is pinned, and RCU held, so the dentries 1285 and inodes won't disappear, neither will our sb or filesystem 1286 module. ->d_sb may be used. 1287 1288 It is a tricky calling convention because it needs to be called 1289 under "rcu-walk", ie. without any locks or references on things. 1290 1291``d_delete`` 1292 called when the last reference to a dentry is dropped and the 1293 dcache is deciding whether or not to cache it. Return 1 to 1294 delete immediately, or 0 to cache the dentry. Default is NULL 1295 which means to always cache a reachable dentry. d_delete must 1296 be constant and idempotent. 1297 1298``d_init`` 1299 called when a dentry is allocated 1300 1301``d_release`` 1302 called when a dentry is really deallocated 1303 1304``d_iput`` 1305 called when a dentry loses its inode (just prior to its being 1306 deallocated). The default when this is NULL is that the VFS 1307 calls iput(). If you define this method, you must call iput() 1308 yourself 1309 1310``d_dname`` 1311 called when the pathname of a dentry should be generated. 1312 Useful for some pseudo filesystems (sockfs, pipefs, ...) to 1313 delay pathname generation. (Instead of doing it when dentry is 1314 created, it's done only when the path is needed.). Real 1315 filesystems probably dont want to use it, because their dentries 1316 are present in global dcache hash, so their hash should be an 1317 invariant. As no lock is held, d_dname() should not try to 1318 modify the dentry itself, unless appropriate SMP safety is used. 1319 CAUTION : d_path() logic is quite tricky. The correct way to 1320 return for example "Hello" is to put it at the end of the 1321 buffer, and returns a pointer to the first char. 1322 dynamic_dname() helper function is provided to take care of 1323 this. 1324 1325 Example : 1326 1327.. code-block:: c 1328 1329 static char *pipefs_dname(struct dentry *dent, char *buffer, int buflen) 1330 { 1331 return dynamic_dname(dentry, buffer, buflen, "pipe:[%lu]", 1332 dentry->d_inode->i_ino); 1333 } 1334 1335``d_automount`` 1336 called when an automount dentry is to be traversed (optional). 1337 This should create a new VFS mount record and return the record 1338 to the caller. The caller is supplied with a path parameter 1339 giving the automount directory to describe the automount target 1340 and the parent VFS mount record to provide inheritable mount 1341 parameters. NULL should be returned if someone else managed to 1342 make the automount first. If the vfsmount creation failed, then 1343 an error code should be returned. If -EISDIR is returned, then 1344 the directory will be treated as an ordinary directory and 1345 returned to pathwalk to continue walking. 1346 1347 If a vfsmount is returned, the caller will attempt to mount it 1348 on the mountpoint and will remove the vfsmount from its 1349 expiration list in the case of failure. The vfsmount should be 1350 returned with 2 refs on it to prevent automatic expiration - the 1351 caller will clean up the additional ref. 1352 1353 This function is only used if DCACHE_NEED_AUTOMOUNT is set on 1354 the dentry. This is set by __d_instantiate() if S_AUTOMOUNT is 1355 set on the inode being added. 1356 1357``d_manage`` 1358 called to allow the filesystem to manage the transition from a 1359 dentry (optional). This allows autofs, for example, to hold up 1360 clients waiting to explore behind a 'mountpoint' while letting 1361 the daemon go past and construct the subtree there. 0 should be 1362 returned to let the calling process continue. -EISDIR can be 1363 returned to tell pathwalk to use this directory as an ordinary 1364 directory and to ignore anything mounted on it and not to check 1365 the automount flag. Any other error code will abort pathwalk 1366 completely. 1367 1368 If the 'rcu_walk' parameter is true, then the caller is doing a 1369 pathwalk in RCU-walk mode. Sleeping is not permitted in this 1370 mode, and the caller can be asked to leave it and call again by 1371 returning -ECHILD. -EISDIR may also be returned to tell 1372 pathwalk to ignore d_automount or any mounts. 1373 1374 This function is only used if DCACHE_MANAGE_TRANSIT is set on 1375 the dentry being transited from. 1376 1377``d_real`` 1378 overlay/union type filesystems implement this method to return 1379 one of the underlying dentries hidden by the overlay. It is 1380 used in two different modes: 1381 1382 Called from file_dentry() it returns the real dentry matching 1383 the inode argument. The real dentry may be from a lower layer 1384 already copied up, but still referenced from the file. This 1385 mode is selected with a non-NULL inode argument. 1386 1387 With NULL inode the topmost real underlying dentry is returned. 1388 1389Each dentry has a pointer to its parent dentry, as well as a hash list 1390of child dentries. Child dentries are basically like files in a 1391directory. 1392 1393 1394Directory Entry Cache API 1395-------------------------- 1396 1397There are a number of functions defined which permit a filesystem to 1398manipulate dentries: 1399 1400``dget`` 1401 open a new handle for an existing dentry (this just increments 1402 the usage count) 1403 1404``dput`` 1405 close a handle for a dentry (decrements the usage count). If 1406 the usage count drops to 0, and the dentry is still in its 1407 parent's hash, the "d_delete" method is called to check whether 1408 it should be cached. If it should not be cached, or if the 1409 dentry is not hashed, it is deleted. Otherwise cached dentries 1410 are put into an LRU list to be reclaimed on memory shortage. 1411 1412``d_drop`` 1413 this unhashes a dentry from its parents hash list. A subsequent 1414 call to dput() will deallocate the dentry if its usage count 1415 drops to 0 1416 1417``d_delete`` 1418 delete a dentry. If there are no other open references to the 1419 dentry then the dentry is turned into a negative dentry (the 1420 d_iput() method is called). If there are other references, then 1421 d_drop() is called instead 1422 1423``d_add`` 1424 add a dentry to its parents hash list and then calls 1425 d_instantiate() 1426 1427``d_instantiate`` 1428 add a dentry to the alias hash list for the inode and updates 1429 the "d_inode" member. The "i_count" member in the inode 1430 structure should be set/incremented. If the inode pointer is 1431 NULL, the dentry is called a "negative dentry". This function 1432 is commonly called when an inode is created for an existing 1433 negative dentry 1434 1435``d_lookup`` 1436 look up a dentry given its parent and path name component It 1437 looks up the child of that given name from the dcache hash 1438 table. If it is found, the reference count is incremented and 1439 the dentry is returned. The caller must use dput() to free the 1440 dentry when it finishes using it. 1441 1442 1443Mount Options 1444============= 1445 1446 1447Parsing options 1448--------------- 1449 1450On mount and remount the filesystem is passed a string containing a 1451comma separated list of mount options. The options can have either of 1452these forms: 1453 1454 option 1455 option=value 1456 1457The <linux/parser.h> header defines an API that helps parse these 1458options. There are plenty of examples on how to use it in existing 1459filesystems. 1460 1461 1462Showing options 1463--------------- 1464 1465If a filesystem accepts mount options, it must define show_options() to 1466show all the currently active options. The rules are: 1467 1468 - options MUST be shown which are not default or their values differ 1469 from the default 1470 1471 - options MAY be shown which are enabled by default or have their 1472 default value 1473 1474Options used only internally between a mount helper and the kernel (such 1475as file descriptors), or which only have an effect during the mounting 1476(such as ones controlling the creation of a journal) are exempt from the 1477above rules. 1478 1479The underlying reason for the above rules is to make sure, that a mount 1480can be accurately replicated (e.g. umounting and mounting again) based 1481on the information found in /proc/mounts. 1482 1483 1484Resources 1485========= 1486 1487(Note some of these resources are not up-to-date with the latest kernel 1488 version.) 1489 1490Creating Linux virtual filesystems. 2002 1491 <https://lwn.net/Articles/13325/> 1492 1493The Linux Virtual File-system Layer by Neil Brown. 1999 1494 <http://www.cse.unsw.edu.au/~neilb/oss/linux-commentary/vfs.html> 1495 1496A tour of the Linux VFS by Michael K. Johnson. 1996 1497 <https://www.tldp.org/LDP/khg/HyperNews/get/fs/vfstour.html> 1498 1499A small trail through the Linux kernel by Andries Brouwer. 2001 1500 <https://www.win.tue.nl/~aeb/linux/vfs/trail.html> 1501