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. As of kernel 2.6.39, 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 struct dentry *(*mount) (struct file_system_type *, int, 119 const char *, void *); 120 void (*kill_sb) (struct super_block *); 121 struct module *owner; 122 struct file_system_type * next; 123 struct list_head fs_supers; 124 struct lock_class_key s_lock_key; 125 struct lock_class_key s_umount_key; 126 }; 127 128``name`` 129 the name of the filesystem type, such as "ext2", "iso9660", 130 "msdos" and so on 131 132``fs_flags`` 133 various flags (i.e. FS_REQUIRES_DEV, FS_NO_DCACHE, etc.) 134 135``mount`` 136 the method to call when a new instance of this filesystem should 137 be mounted 138 139``kill_sb`` 140 the method to call when an instance of this filesystem should be 141 shut down 142 143 144``owner`` 145 for internal VFS use: you should initialize this to THIS_MODULE 146 in most cases. 147 148``next`` 149 for internal VFS use: you should initialize this to NULL 150 151 s_lock_key, s_umount_key: lockdep-specific 152 153The mount() method has the following arguments: 154 155``struct file_system_type *fs_type`` 156 describes the filesystem, partly initialized by the specific 157 filesystem code 158 159``int flags`` 160 mount flags 161 162``const char *dev_name`` 163 the device name we are mounting. 164 165``void *data`` 166 arbitrary mount options, usually comes as an ASCII string (see 167 "Mount Options" section) 168 169The mount() method must return the root dentry of the tree requested by 170caller. An active reference to its superblock must be grabbed and the 171superblock must be locked. On failure it should return ERR_PTR(error). 172 173The arguments match those of mount(2) and their interpretation depends 174on filesystem type. E.g. for block filesystems, dev_name is interpreted 175as block device name, that device is opened and if it contains a 176suitable filesystem image the method creates and initializes struct 177super_block accordingly, returning its root dentry to caller. 178 179->mount() may choose to return a subtree of existing filesystem - it 180doesn't have to create a new one. The main result from the caller's 181point of view is a reference to dentry at the root of (sub)tree to be 182attached; creation of new superblock is a common side effect. 183 184The most interesting member of the superblock structure that the mount() 185method fills in is the "s_op" field. This is a pointer to a "struct 186super_operations" which describes the next level of the filesystem 187implementation. 188 189Usually, a filesystem uses one of the generic mount() implementations 190and provides a fill_super() callback instead. The generic variants are: 191 192``mount_bdev`` 193 mount a filesystem residing on a block device 194 195``mount_nodev`` 196 mount a filesystem that is not backed by a device 197 198``mount_single`` 199 mount a filesystem which shares the instance between all mounts 200 201A fill_super() callback implementation has the following arguments: 202 203``struct super_block *sb`` 204 the superblock structure. The callback must initialize this 205 properly. 206 207``void *data`` 208 arbitrary mount options, usually comes as an ASCII string (see 209 "Mount Options" section) 210 211``int silent`` 212 whether or not to be silent on error 213 214 215The Superblock Object 216===================== 217 218A superblock object represents a mounted filesystem. 219 220 221struct super_operations 222----------------------- 223 224This describes how the VFS can manipulate the superblock of your 225filesystem. As of kernel 2.6.22, the following members are defined: 226 227.. code-block:: c 228 229 struct super_operations { 230 struct inode *(*alloc_inode)(struct super_block *sb); 231 void (*destroy_inode)(struct inode *); 232 233 void (*dirty_inode) (struct inode *, int flags); 234 int (*write_inode) (struct inode *, int); 235 void (*drop_inode) (struct inode *); 236 void (*delete_inode) (struct inode *); 237 void (*put_super) (struct super_block *); 238 int (*sync_fs)(struct super_block *sb, int wait); 239 int (*freeze_fs) (struct super_block *); 240 int (*unfreeze_fs) (struct super_block *); 241 int (*statfs) (struct dentry *, struct kstatfs *); 242 int (*remount_fs) (struct super_block *, int *, char *); 243 void (*clear_inode) (struct inode *); 244 void (*umount_begin) (struct super_block *); 245 246 int (*show_options)(struct seq_file *, struct dentry *); 247 248 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t); 249 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t); 250 int (*nr_cached_objects)(struct super_block *); 251 void (*free_cached_objects)(struct super_block *, int); 252 }; 253 254All methods are called without any locks being held, unless otherwise 255noted. This means that most methods can block safely. All methods are 256only called from a process context (i.e. not from an interrupt handler 257or bottom half). 258 259``alloc_inode`` 260 this method is called by alloc_inode() to allocate memory for 261 struct inode and initialize it. If this function is not 262 defined, a simple 'struct inode' is allocated. Normally 263 alloc_inode will be used to allocate a larger structure which 264 contains a 'struct inode' embedded within it. 265 266``destroy_inode`` 267 this method is called by destroy_inode() to release resources 268 allocated for struct inode. It is only required if 269 ->alloc_inode was defined and simply undoes anything done by 270 ->alloc_inode. 271 272``dirty_inode`` 273 this method is called by the VFS when an inode is marked dirty. 274 This is specifically for the inode itself being marked dirty, 275 not its data. If the update needs to be persisted by fdatasync(), 276 then I_DIRTY_DATASYNC will be set in the flags argument. 277 I_DIRTY_TIME will be set in the flags in case lazytime is enabled 278 and struct inode has times updated since the last ->dirty_inode 279 call. 280 281``write_inode`` 282 this method is called when the VFS needs to write an inode to 283 disc. The second parameter indicates whether the write should 284 be synchronous or not, not all filesystems check this flag. 285 286``drop_inode`` 287 called when the last access to the inode is dropped, with the 288 inode->i_lock spinlock held. 289 290 This method should be either NULL (normal UNIX filesystem 291 semantics) or "generic_delete_inode" (for filesystems that do 292 not want to cache inodes - causing "delete_inode" to always be 293 called regardless of the value of i_nlink) 294 295 The "generic_delete_inode()" behavior is equivalent to the old 296 practice of using "force_delete" in the put_inode() case, but 297 does not have the races that the "force_delete()" approach had. 298 299``delete_inode`` 300 called when the VFS wants to delete an inode 301 302``put_super`` 303 called when the VFS wishes to free the superblock 304 (i.e. unmount). This is called with the superblock lock held 305 306``sync_fs`` 307 called when VFS is writing out all dirty data associated with a 308 superblock. The second parameter indicates whether the method 309 should wait until the write out has been completed. Optional. 310 311``freeze_fs`` 312 called when VFS is locking a filesystem and forcing it into a 313 consistent state. This method is currently used by the Logical 314 Volume Manager (LVM). 315 316``unfreeze_fs`` 317 called when VFS is unlocking a filesystem and making it writable 318 again. 319 320``statfs`` 321 called when the VFS needs to get filesystem statistics. 322 323``remount_fs`` 324 called when the filesystem is remounted. This is called with 325 the kernel lock held 326 327``clear_inode`` 328 called then the VFS clears the inode. Optional 329 330``umount_begin`` 331 called when the VFS is unmounting a filesystem. 332 333``show_options`` 334 called by the VFS to show mount options for /proc/<pid>/mounts. 335 (see "Mount Options" section) 336 337``quota_read`` 338 called by the VFS to read from filesystem quota file. 339 340``quota_write`` 341 called by the VFS to write to filesystem quota file. 342 343``nr_cached_objects`` 344 called by the sb cache shrinking function for the filesystem to 345 return the number of freeable cached objects it contains. 346 Optional. 347 348``free_cache_objects`` 349 called by the sb cache shrinking function for the filesystem to 350 scan the number of objects indicated to try to free them. 351 Optional, but any filesystem implementing this method needs to 352 also implement ->nr_cached_objects for it to be called 353 correctly. 354 355 We can't do anything with any errors that the filesystem might 356 encountered, hence the void return type. This will never be 357 called if the VM is trying to reclaim under GFP_NOFS conditions, 358 hence this method does not need to handle that situation itself. 359 360 Implementations must include conditional reschedule calls inside 361 any scanning loop that is done. This allows the VFS to 362 determine appropriate scan batch sizes without having to worry 363 about whether implementations will cause holdoff problems due to 364 large scan batch sizes. 365 366Whoever sets up the inode is responsible for filling in the "i_op" 367field. This is a pointer to a "struct inode_operations" which describes 368the methods that can be performed on individual inodes. 369 370 371struct xattr_handlers 372--------------------- 373 374On filesystems that support extended attributes (xattrs), the s_xattr 375superblock field points to a NULL-terminated array of xattr handlers. 376Extended attributes are name:value pairs. 377 378``name`` 379 Indicates that the handler matches attributes with the specified 380 name (such as "system.posix_acl_access"); the prefix field must 381 be NULL. 382 383``prefix`` 384 Indicates that the handler matches all attributes with the 385 specified name prefix (such as "user."); the name field must be 386 NULL. 387 388``list`` 389 Determine if attributes matching this xattr handler should be 390 listed for a particular dentry. Used by some listxattr 391 implementations like generic_listxattr. 392 393``get`` 394 Called by the VFS to get the value of a particular extended 395 attribute. This method is called by the getxattr(2) system 396 call. 397 398``set`` 399 Called by the VFS to set the value of a particular extended 400 attribute. When the new value is NULL, called to remove a 401 particular extended attribute. This method is called by the 402 setxattr(2) and removexattr(2) system calls. 403 404When none of the xattr handlers of a filesystem match the specified 405attribute name or when a filesystem doesn't support extended attributes, 406the various ``*xattr(2)`` system calls return -EOPNOTSUPP. 407 408 409The Inode Object 410================ 411 412An inode object represents an object within the filesystem. 413 414 415struct inode_operations 416----------------------- 417 418This describes how the VFS can manipulate an inode in your filesystem. 419As of kernel 2.6.22, the following members are defined: 420 421.. code-block:: c 422 423 struct inode_operations { 424 int (*create) (struct user_namespace *, struct inode *,struct dentry *, umode_t, bool); 425 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int); 426 int (*link) (struct dentry *,struct inode *,struct dentry *); 427 int (*unlink) (struct inode *,struct dentry *); 428 int (*symlink) (struct user_namespace *, struct inode *,struct dentry *,const char *); 429 int (*mkdir) (struct user_namespace *, struct inode *,struct dentry *,umode_t); 430 int (*rmdir) (struct inode *,struct dentry *); 431 int (*mknod) (struct user_namespace *, struct inode *,struct dentry *,umode_t,dev_t); 432 int (*rename) (struct user_namespace *, struct inode *, struct dentry *, 433 struct inode *, struct dentry *, unsigned int); 434 int (*readlink) (struct dentry *, char __user *,int); 435 const char *(*get_link) (struct dentry *, struct inode *, 436 struct delayed_call *); 437 int (*permission) (struct user_namespace *, struct inode *, int); 438 struct posix_acl * (*get_inode_acl)(struct inode *, int, bool); 439 int (*setattr) (struct user_namespace *, struct dentry *, struct iattr *); 440 int (*getattr) (struct user_namespace *, const struct path *, struct kstat *, u32, unsigned int); 441 ssize_t (*listxattr) (struct dentry *, char *, size_t); 442 void (*update_time)(struct inode *, struct timespec *, int); 443 int (*atomic_open)(struct inode *, struct dentry *, struct file *, 444 unsigned open_flag, umode_t create_mode); 445 int (*tmpfile) (struct user_namespace *, struct inode *, struct file *, umode_t); 446 struct posix_acl * (*get_acl)(struct user_namespace *, struct dentry *, int); 447 int (*set_acl)(struct user_namespace *, struct dentry *, struct posix_acl *, int); 448 int (*fileattr_set)(struct user_namespace *mnt_userns, 449 struct dentry *dentry, struct fileattr *fa); 450 int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa); 451 }; 452 453Again, all methods are called without any locks being held, unless 454otherwise noted. 455 456``create`` 457 called by the open(2) and creat(2) system calls. Only required 458 if you want to support regular files. The dentry you get should 459 not have an inode (i.e. it should be a negative dentry). Here 460 you will probably call d_instantiate() with the dentry and the 461 newly created inode 462 463``lookup`` 464 called when the VFS needs to look up an inode in a parent 465 directory. The name to look for is found in the dentry. This 466 method must call d_add() to insert the found inode into the 467 dentry. The "i_count" field in the inode structure should be 468 incremented. If the named inode does not exist a NULL inode 469 should be inserted into the dentry (this is called a negative 470 dentry). Returning an error code from this routine must only be 471 done on a real error, otherwise creating inodes with system 472 calls like create(2), mknod(2), mkdir(2) and so on will fail. 473 If you wish to overload the dentry methods then you should 474 initialise the "d_dop" field in the dentry; this is a pointer to 475 a struct "dentry_operations". This method is called with the 476 directory inode semaphore held 477 478``link`` 479 called by the link(2) system call. Only required if you want to 480 support hard links. You will probably need to call 481 d_instantiate() just as you would in the create() method 482 483``unlink`` 484 called by the unlink(2) system call. Only required if you want 485 to support deleting inodes 486 487``symlink`` 488 called by the symlink(2) system call. Only required if you want 489 to support symlinks. You will probably need to call 490 d_instantiate() just as you would in the create() method 491 492``mkdir`` 493 called by the mkdir(2) system call. Only required if you want 494 to support creating subdirectories. You will probably need to 495 call d_instantiate() just as you would in the create() method 496 497``rmdir`` 498 called by the rmdir(2) system call. Only required if you want 499 to support deleting subdirectories 500 501``mknod`` 502 called by the mknod(2) system call to create a device (char, 503 block) inode or a named pipe (FIFO) or socket. Only required if 504 you want to support creating these types of inodes. You will 505 probably need to call d_instantiate() just as you would in the 506 create() method 507 508``rename`` 509 called by the rename(2) system call to rename the object to have 510 the parent and name given by the second inode and dentry. 511 512 The filesystem must return -EINVAL for any unsupported or 513 unknown flags. Currently the following flags are implemented: 514 (1) RENAME_NOREPLACE: this flag indicates that if the target of 515 the rename exists the rename should fail with -EEXIST instead of 516 replacing the target. The VFS already checks for existence, so 517 for local filesystems the RENAME_NOREPLACE implementation is 518 equivalent to plain rename. 519 (2) RENAME_EXCHANGE: exchange source and target. Both must 520 exist; this is checked by the VFS. Unlike plain rename, source 521 and target may be of different type. 522 523``get_link`` 524 called by the VFS to follow a symbolic link to the inode it 525 points to. Only required if you want to support symbolic links. 526 This method returns the symlink body to traverse (and possibly 527 resets the current position with nd_jump_link()). If the body 528 won't go away until the inode is gone, nothing else is needed; 529 if it needs to be otherwise pinned, arrange for its release by 530 having get_link(..., ..., done) do set_delayed_call(done, 531 destructor, argument). In that case destructor(argument) will 532 be called once VFS is done with the body you've returned. May 533 be called in RCU mode; that is indicated by NULL dentry 534 argument. If request can't be handled without leaving RCU mode, 535 have it return ERR_PTR(-ECHILD). 536 537 If the filesystem stores the symlink target in ->i_link, the 538 VFS may use it directly without calling ->get_link(); however, 539 ->get_link() must still be provided. ->i_link must not be 540 freed until after an RCU grace period. Writing to ->i_link 541 post-iget() time requires a 'release' memory barrier. 542 543``readlink`` 544 this is now just an override for use by readlink(2) for the 545 cases when ->get_link uses nd_jump_link() or object is not in 546 fact a symlink. Normally filesystems should only implement 547 ->get_link for symlinks and readlink(2) will automatically use 548 that. 549 550``permission`` 551 called by the VFS to check for access rights on a POSIX-like 552 filesystem. 553 554 May be called in rcu-walk mode (mask & MAY_NOT_BLOCK). If in 555 rcu-walk mode, the filesystem must check the permission without 556 blocking or storing to the inode. 557 558 If a situation is encountered that rcu-walk cannot handle, 559 return 560 -ECHILD and it will be called again in ref-walk mode. 561 562``setattr`` 563 called by the VFS to set attributes for a file. This method is 564 called by chmod(2) and related system calls. 565 566``getattr`` 567 called by the VFS to get attributes of a file. This method is 568 called by stat(2) and related system calls. 569 570``listxattr`` 571 called by the VFS to list all extended attributes for a given 572 file. This method is called by the listxattr(2) system call. 573 574``update_time`` 575 called by the VFS to update a specific time or the i_version of 576 an inode. If this is not defined the VFS will update the inode 577 itself and call mark_inode_dirty_sync. 578 579``atomic_open`` 580 called on the last component of an open. Using this optional 581 method the filesystem can look up, possibly create and open the 582 file in one atomic operation. If it wants to leave actual 583 opening to the caller (e.g. if the file turned out to be a 584 symlink, device, or just something filesystem won't do atomic 585 open for), it may signal this by returning finish_no_open(file, 586 dentry). This method is only called if the last component is 587 negative or needs lookup. Cached positive dentries are still 588 handled by f_op->open(). If the file was created, FMODE_CREATED 589 flag should be set in file->f_mode. In case of O_EXCL the 590 method must only succeed if the file didn't exist and hence 591 FMODE_CREATED shall always be set on success. 592 593``tmpfile`` 594 called in the end of O_TMPFILE open(). Optional, equivalent to 595 atomically creating, opening and unlinking a file in given 596 directory. On success needs to return with the file already 597 open; this can be done by calling finish_open_simple() right at 598 the end. 599 600``fileattr_get`` 601 called on ioctl(FS_IOC_GETFLAGS) and ioctl(FS_IOC_FSGETXATTR) to 602 retrieve miscellaneous file flags and attributes. Also called 603 before the relevant SET operation to check what is being changed 604 (in this case with i_rwsem locked exclusive). If unset, then 605 fall back to f_op->ioctl(). 606 607``fileattr_set`` 608 called on ioctl(FS_IOC_SETFLAGS) and ioctl(FS_IOC_FSSETXATTR) to 609 change miscellaneous file flags and attributes. Callers hold 610 i_rwsem exclusive. If unset, then fall back to f_op->ioctl(). 611 612 613The Address Space Object 614======================== 615 616The address space object is used to group and manage pages in the page 617cache. It can be used to keep track of the pages in a file (or anything 618else) and also track the mapping of sections of the file into process 619address spaces. 620 621There are a number of distinct yet related services that an 622address-space can provide. These include communicating memory pressure, 623page lookup by address, and keeping track of pages tagged as Dirty or 624Writeback. 625 626The first can be used independently to the others. The VM can try to 627either write dirty pages in order to clean them, or release clean pages 628in order to reuse them. To do this it can call the ->writepage method 629on dirty pages, and ->release_folio on clean folios with the private 630flag set. Clean pages without PagePrivate and with no external references 631will be released without notice being given to the address_space. 632 633To achieve this functionality, pages need to be placed on an LRU with 634lru_cache_add and mark_page_active needs to be called whenever the page 635is used. 636 637Pages are normally kept in a radix tree index by ->index. This tree 638maintains information about the PG_Dirty and PG_Writeback status of each 639page, so that pages with either of these flags can be found quickly. 640 641The Dirty tag is primarily used by mpage_writepages - the default 642->writepages method. It uses the tag to find dirty pages to call 643->writepage on. If mpage_writepages is not used (i.e. the address 644provides its own ->writepages) , the PAGECACHE_TAG_DIRTY tag is almost 645unused. write_inode_now and sync_inode do use it (through 646__sync_single_inode) to check if ->writepages has been successful in 647writing out the whole address_space. 648 649The Writeback tag is used by filemap*wait* and sync_page* functions, via 650filemap_fdatawait_range, to wait for all writeback to complete. 651 652An address_space handler may attach extra information to a page, 653typically using the 'private' field in the 'struct page'. If such 654information is attached, the PG_Private flag should be set. This will 655cause various VM routines to make extra calls into the address_space 656handler to deal with that data. 657 658An address space acts as an intermediate between storage and 659application. Data is read into the address space a whole page at a 660time, and provided to the application either by copying of the page, or 661by memory-mapping the page. Data is written into the address space by 662the application, and then written-back to storage typically in whole 663pages, however the address_space has finer control of write sizes. 664 665The read process essentially only requires 'read_folio'. The write 666process is more complicated and uses write_begin/write_end or 667dirty_folio to write data into the address_space, and writepage and 668writepages to writeback data to storage. 669 670Adding and removing pages to/from an address_space is protected by the 671inode's i_mutex. 672 673When data is written to a page, the PG_Dirty flag should be set. It 674typically remains set until writepage asks for it to be written. This 675should clear PG_Dirty and set PG_Writeback. It can be actually written 676at any point after PG_Dirty is clear. Once it is known to be safe, 677PG_Writeback is cleared. 678 679Writeback makes use of a writeback_control structure to direct the 680operations. This gives the writepage and writepages operations some 681information about the nature of and reason for the writeback request, 682and the constraints under which it is being done. It is also used to 683return information back to the caller about the result of a writepage or 684writepages request. 685 686 687Handling errors during writeback 688-------------------------------- 689 690Most applications that do buffered I/O will periodically call a file 691synchronization call (fsync, fdatasync, msync or sync_file_range) to 692ensure that data written has made it to the backing store. When there 693is an error during writeback, they expect that error to be reported when 694a file sync request is made. After an error has been reported on one 695request, subsequent requests on the same file descriptor should return 6960, unless further writeback errors have occurred since the previous file 697syncronization. 698 699Ideally, the kernel would report errors only on file descriptions on 700which writes were done that subsequently failed to be written back. The 701generic pagecache infrastructure does not track the file descriptions 702that have dirtied each individual page however, so determining which 703file descriptors should get back an error is not possible. 704 705Instead, the generic writeback error tracking infrastructure in the 706kernel settles for reporting errors to fsync on all file descriptions 707that were open at the time that the error occurred. In a situation with 708multiple writers, all of them will get back an error on a subsequent 709fsync, even if all of the writes done through that particular file 710descriptor succeeded (or even if there were no writes on that file 711descriptor at all). 712 713Filesystems that wish to use this infrastructure should call 714mapping_set_error to record the error in the address_space when it 715occurs. Then, after writing back data from the pagecache in their 716file->fsync operation, they should call file_check_and_advance_wb_err to 717ensure that the struct file's error cursor has advanced to the correct 718point in the stream of errors emitted by the backing device(s). 719 720 721struct address_space_operations 722------------------------------- 723 724This describes how the VFS can manipulate mapping of a file to page 725cache in your filesystem. The following members are defined: 726 727.. code-block:: c 728 729 struct address_space_operations { 730 int (*writepage)(struct page *page, struct writeback_control *wbc); 731 int (*read_folio)(struct file *, struct folio *); 732 int (*writepages)(struct address_space *, struct writeback_control *); 733 bool (*dirty_folio)(struct address_space *, struct folio *); 734 void (*readahead)(struct readahead_control *); 735 int (*write_begin)(struct file *, struct address_space *mapping, 736 loff_t pos, unsigned len, 737 struct page **pagep, void **fsdata); 738 int (*write_end)(struct file *, struct address_space *mapping, 739 loff_t pos, unsigned len, unsigned copied, 740 struct page *page, void *fsdata); 741 sector_t (*bmap)(struct address_space *, sector_t); 742 void (*invalidate_folio) (struct folio *, size_t start, size_t len); 743 bool (*release_folio)(struct folio *, gfp_t); 744 void (*free_folio)(struct folio *); 745 ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter); 746 int (*migrate_folio)(struct mapping *, struct folio *dst, 747 struct folio *src, enum migrate_mode); 748 int (*launder_folio) (struct folio *); 749 750 bool (*is_partially_uptodate) (struct folio *, size_t from, 751 size_t count); 752 void (*is_dirty_writeback)(struct folio *, bool *, bool *); 753 int (*error_remove_page) (struct mapping *mapping, struct page *page); 754 int (*swap_activate)(struct swap_info_struct *sis, struct file *f, sector_t *span) 755 int (*swap_deactivate)(struct file *); 756 int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter); 757 }; 758 759``writepage`` 760 called by the VM to write a dirty page to backing store. This 761 may happen for data integrity reasons (i.e. 'sync'), or to free 762 up memory (flush). The difference can be seen in 763 wbc->sync_mode. The PG_Dirty flag has been cleared and 764 PageLocked is true. writepage should start writeout, should set 765 PG_Writeback, and should make sure the page is unlocked, either 766 synchronously or asynchronously when the write operation 767 completes. 768 769 If wbc->sync_mode is WB_SYNC_NONE, ->writepage doesn't have to 770 try too hard if there are problems, and may choose to write out 771 other pages from the mapping if that is easier (e.g. due to 772 internal dependencies). If it chooses not to start writeout, it 773 should return AOP_WRITEPAGE_ACTIVATE so that the VM will not 774 keep calling ->writepage on that page. 775 776 See the file "Locking" for more details. 777 778``read_folio`` 779 Called by the page cache to read a folio from the backing store. 780 The 'file' argument supplies authentication information to network 781 filesystems, and is generally not used by block based filesystems. 782 It may be NULL if the caller does not have an open file (eg if 783 the kernel is performing a read for itself rather than on behalf 784 of a userspace process with an open file). 785 786 If the mapping does not support large folios, the folio will 787 contain a single page. The folio will be locked when read_folio 788 is called. If the read completes successfully, the folio should 789 be marked uptodate. The filesystem should unlock the folio 790 once the read has completed, whether it was successful or not. 791 The filesystem does not need to modify the refcount on the folio; 792 the page cache holds a reference count and that will not be 793 released until the folio is unlocked. 794 795 Filesystems may implement ->read_folio() synchronously. 796 In normal operation, folios are read through the ->readahead() 797 method. Only if this fails, or if the caller needs to wait for 798 the read to complete will the page cache call ->read_folio(). 799 Filesystems should not attempt to perform their own readahead 800 in the ->read_folio() operation. 801 802 If the filesystem cannot perform the read at this time, it can 803 unlock the folio, do whatever action it needs to ensure that the 804 read will succeed in the future and return AOP_TRUNCATED_PAGE. 805 In this case, the caller should look up the folio, lock it, 806 and call ->read_folio again. 807 808 Callers may invoke the ->read_folio() method directly, but using 809 read_mapping_folio() will take care of locking, waiting for the 810 read to complete and handle cases such as AOP_TRUNCATED_PAGE. 811 812``writepages`` 813 called by the VM to write out pages associated with the 814 address_space object. If wbc->sync_mode is WB_SYNC_ALL, then 815 the writeback_control will specify a range of pages that must be 816 written out. If it is WB_SYNC_NONE, then a nr_to_write is 817 given and that many pages should be written if possible. If no 818 ->writepages is given, then mpage_writepages is used instead. 819 This will choose pages from the address space that are tagged as 820 DIRTY and will pass them to ->writepage. 821 822``dirty_folio`` 823 called by the VM to mark a folio as dirty. This is particularly 824 needed if an address space attaches private data to a folio, and 825 that data needs to be updated when a folio is dirtied. This is 826 called, for example, when a memory mapped page gets modified. 827 If defined, it should set the folio dirty flag, and the 828 PAGECACHE_TAG_DIRTY search mark in i_pages. 829 830``readahead`` 831 Called by the VM to read pages associated with the address_space 832 object. The pages are consecutive in the page cache and are 833 locked. The implementation should decrement the page refcount 834 after starting I/O on each page. Usually the page will be 835 unlocked by the I/O completion handler. The set of pages are 836 divided into some sync pages followed by some async pages, 837 rac->ra->async_size gives the number of async pages. The 838 filesystem should attempt to read all sync pages but may decide 839 to stop once it reaches the async pages. If it does decide to 840 stop attempting I/O, it can simply return. The caller will 841 remove the remaining pages from the address space, unlock them 842 and decrement the page refcount. Set PageUptodate if the I/O 843 completes successfully. Setting PageError on any page will be 844 ignored; simply unlock the page if an I/O error occurs. 845 846``write_begin`` 847 Called by the generic buffered write code to ask the filesystem 848 to prepare to write len bytes at the given offset in the file. 849 The address_space should check that the write will be able to 850 complete, by allocating space if necessary and doing any other 851 internal housekeeping. If the write will update parts of any 852 basic-blocks on storage, then those blocks should be pre-read 853 (if they haven't been read already) so that the updated blocks 854 can be written out properly. 855 856 The filesystem must return the locked pagecache page for the 857 specified offset, in ``*pagep``, for the caller to write into. 858 859 It must be able to cope with short writes (where the length 860 passed to write_begin is greater than the number of bytes copied 861 into the page). 862 863 A void * may be returned in fsdata, which then gets passed into 864 write_end. 865 866 Returns 0 on success; < 0 on failure (which is the error code), 867 in which case write_end is not called. 868 869``write_end`` 870 After a successful write_begin, and data copy, write_end must be 871 called. len is the original len passed to write_begin, and 872 copied is the amount that was able to be copied. 873 874 The filesystem must take care of unlocking the page and 875 releasing it refcount, and updating i_size. 876 877 Returns < 0 on failure, otherwise the number of bytes (<= 878 'copied') that were able to be copied into pagecache. 879 880``bmap`` 881 called by the VFS to map a logical block offset within object to 882 physical block number. This method is used by the FIBMAP ioctl 883 and for working with swap-files. To be able to swap to a file, 884 the file must have a stable mapping to a block device. The swap 885 system does not go through the filesystem but instead uses bmap 886 to find out where the blocks in the file are and uses those 887 addresses directly. 888 889``invalidate_folio`` 890 If a folio has private data, then invalidate_folio will be 891 called when part or all of the folio is to be removed from the 892 address space. This generally corresponds to either a 893 truncation, punch hole or a complete invalidation of the address 894 space (in the latter case 'offset' will always be 0 and 'length' 895 will be folio_size()). Any private data associated with the folio 896 should be updated to reflect this truncation. If offset is 0 897 and length is folio_size(), then the private data should be 898 released, because the folio must be able to be completely 899 discarded. This may be done by calling the ->release_folio 900 function, but in this case the release MUST succeed. 901 902``release_folio`` 903 release_folio is called on folios with private data to tell the 904 filesystem that the folio is about to be freed. ->release_folio 905 should remove any private data from the folio and clear the 906 private flag. If release_folio() fails, it should return false. 907 release_folio() is used in two distinct though related cases. 908 The first is when the VM wants to free a clean folio with no 909 active users. If ->release_folio succeeds, the folio will be 910 removed from the address_space and be freed. 911 912 The second case is when a request has been made to invalidate 913 some or all folios in an address_space. This can happen 914 through the fadvise(POSIX_FADV_DONTNEED) system call or by the 915 filesystem explicitly requesting it as nfs and 9p do (when they 916 believe the cache may be out of date with storage) by calling 917 invalidate_inode_pages2(). If the filesystem makes such a call, 918 and needs to be certain that all folios are invalidated, then 919 its release_folio will need to ensure this. Possibly it can 920 clear the uptodate flag if it cannot free private data yet. 921 922``free_folio`` 923 free_folio is called once the folio is no longer visible in the 924 page cache in order to allow the cleanup of any private data. 925 Since it may be called by the memory reclaimer, it should not 926 assume that the original address_space mapping still exists, and 927 it should not block. 928 929``direct_IO`` 930 called by the generic read/write routines to perform direct_IO - 931 that is IO requests which bypass the page cache and transfer 932 data directly between the storage and the application's address 933 space. 934 935``migrate_folio`` 936 This is used to compact the physical memory usage. If the VM 937 wants to relocate a folio (maybe from a memory device that is 938 signalling imminent failure) it will pass a new folio and an old 939 folio to this function. migrate_folio should transfer any private 940 data across and update any references that it has to the folio. 941 942``launder_folio`` 943 Called before freeing a folio - it writes back the dirty folio. 944 To prevent redirtying the folio, it is kept locked during the 945 whole operation. 946 947``is_partially_uptodate`` 948 Called by the VM when reading a file through the pagecache when 949 the underlying blocksize is smaller than the size of the folio. 950 If the required block is up to date then the read can complete 951 without needing I/O to bring the whole page up to date. 952 953``is_dirty_writeback`` 954 Called by the VM when attempting to reclaim a folio. The VM uses 955 dirty and writeback information to determine if it needs to 956 stall to allow flushers a chance to complete some IO. 957 Ordinarily it can use folio_test_dirty and folio_test_writeback but 958 some filesystems have more complex state (unstable folios in NFS 959 prevent reclaim) or do not set those flags due to locking 960 problems. This callback allows a filesystem to indicate to the 961 VM if a folio should be treated as dirty or writeback for the 962 purposes of stalling. 963 964``error_remove_page`` 965 normally set to generic_error_remove_page if truncation is ok 966 for this address space. Used for memory failure handling. 967 Setting this implies you deal with pages going away under you, 968 unless you have them locked or reference counts increased. 969 970``swap_activate`` 971 972 Called to prepare the given file for swap. It should perform 973 any validation and preparation necessary to ensure that writes 974 can be performed with minimal memory allocation. It should call 975 add_swap_extent(), or the helper iomap_swapfile_activate(), and 976 return the number of extents added. If IO should be submitted 977 through ->swap_rw(), it should set SWP_FS_OPS, otherwise IO will 978 be submitted directly to the block device ``sis->bdev``. 979 980``swap_deactivate`` 981 Called during swapoff on files where swap_activate was 982 successful. 983 984``swap_rw`` 985 Called to read or write swap pages when SWP_FS_OPS is set. 986 987The File Object 988=============== 989 990A file object represents a file opened by a process. This is also known 991as an "open file description" in POSIX parlance. 992 993 994struct file_operations 995---------------------- 996 997This describes how the VFS can manipulate an open file. As of kernel 9984.18, the following members are defined: 999 1000.. code-block:: c 1001 1002 struct file_operations { 1003 struct module *owner; 1004 loff_t (*llseek) (struct file *, loff_t, int); 1005 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *); 1006 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *); 1007 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *); 1008 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *); 1009 int (*iopoll)(struct kiocb *kiocb, bool spin); 1010 int (*iterate) (struct file *, struct dir_context *); 1011 int (*iterate_shared) (struct file *, struct dir_context *); 1012 __poll_t (*poll) (struct file *, struct poll_table_struct *); 1013 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long); 1014 long (*compat_ioctl) (struct file *, unsigned int, unsigned long); 1015 int (*mmap) (struct file *, struct vm_area_struct *); 1016 int (*open) (struct inode *, struct file *); 1017 int (*flush) (struct file *, fl_owner_t id); 1018 int (*release) (struct inode *, struct file *); 1019 int (*fsync) (struct file *, loff_t, loff_t, int datasync); 1020 int (*fasync) (int, struct file *, int); 1021 int (*lock) (struct file *, int, struct file_lock *); 1022 ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int); 1023 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 1024 int (*check_flags)(int); 1025 int (*flock) (struct file *, int, struct file_lock *); 1026 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int); 1027 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int); 1028 int (*setlease)(struct file *, long, struct file_lock **, void **); 1029 long (*fallocate)(struct file *file, int mode, loff_t offset, 1030 loff_t len); 1031 void (*show_fdinfo)(struct seq_file *m, struct file *f); 1032 #ifndef CONFIG_MMU 1033 unsigned (*mmap_capabilities)(struct file *); 1034 #endif 1035 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *, loff_t, size_t, unsigned int); 1036 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in, 1037 struct file *file_out, loff_t pos_out, 1038 loff_t len, unsigned int remap_flags); 1039 int (*fadvise)(struct file *, loff_t, loff_t, int); 1040 }; 1041 1042Again, all methods are called without any locks being held, unless 1043otherwise noted. 1044 1045``llseek`` 1046 called when the VFS needs to move the file position index 1047 1048``read`` 1049 called by read(2) and related system calls 1050 1051``read_iter`` 1052 possibly asynchronous read with iov_iter as destination 1053 1054``write`` 1055 called by write(2) and related system calls 1056 1057``write_iter`` 1058 possibly asynchronous write with iov_iter as source 1059 1060``iopoll`` 1061 called when aio wants to poll for completions on HIPRI iocbs 1062 1063``iterate`` 1064 called when the VFS needs to read the directory contents 1065 1066``iterate_shared`` 1067 called when the VFS needs to read the directory contents when 1068 filesystem supports concurrent dir iterators 1069 1070``poll`` 1071 called by the VFS when a process wants to check if there is 1072 activity on this file and (optionally) go to sleep until there 1073 is activity. Called by the select(2) and poll(2) system calls 1074 1075``unlocked_ioctl`` 1076 called by the ioctl(2) system call. 1077 1078``compat_ioctl`` 1079 called by the ioctl(2) system call when 32 bit system calls are 1080 used on 64 bit kernels. 1081 1082``mmap`` 1083 called by the mmap(2) system call 1084 1085``open`` 1086 called by the VFS when an inode should be opened. When the VFS 1087 opens a file, it creates a new "struct file". It then calls the 1088 open method for the newly allocated file structure. You might 1089 think that the open method really belongs in "struct 1090 inode_operations", and you may be right. I think it's done the 1091 way it is because it makes filesystems simpler to implement. 1092 The open() method is a good place to initialize the 1093 "private_data" member in the file structure if you want to point 1094 to a device structure 1095 1096``flush`` 1097 called by the close(2) system call to flush a file 1098 1099``release`` 1100 called when the last reference to an open file is closed 1101 1102``fsync`` 1103 called by the fsync(2) system call. Also see the section above 1104 entitled "Handling errors during writeback". 1105 1106``fasync`` 1107 called by the fcntl(2) system call when asynchronous 1108 (non-blocking) mode is enabled for a file 1109 1110``lock`` 1111 called by the fcntl(2) system call for F_GETLK, F_SETLK, and 1112 F_SETLKW commands 1113 1114``get_unmapped_area`` 1115 called by the mmap(2) system call 1116 1117``check_flags`` 1118 called by the fcntl(2) system call for F_SETFL command 1119 1120``flock`` 1121 called by the flock(2) system call 1122 1123``splice_write`` 1124 called by the VFS to splice data from a pipe to a file. This 1125 method is used by the splice(2) system call 1126 1127``splice_read`` 1128 called by the VFS to splice data from file to a pipe. This 1129 method is used by the splice(2) system call 1130 1131``setlease`` 1132 called by the VFS to set or release a file lock lease. setlease 1133 implementations should call generic_setlease to record or remove 1134 the lease in the inode after setting it. 1135 1136``fallocate`` 1137 called by the VFS to preallocate blocks or punch a hole. 1138 1139``copy_file_range`` 1140 called by the copy_file_range(2) system call. 1141 1142``remap_file_range`` 1143 called by the ioctl(2) system call for FICLONERANGE and FICLONE 1144 and FIDEDUPERANGE commands to remap file ranges. An 1145 implementation should remap len bytes at pos_in of the source 1146 file into the dest file at pos_out. Implementations must handle 1147 callers passing in len == 0; this means "remap to the end of the 1148 source file". The return value should the number of bytes 1149 remapped, or the usual negative error code if errors occurred 1150 before any bytes were remapped. The remap_flags parameter 1151 accepts REMAP_FILE_* flags. If REMAP_FILE_DEDUP is set then the 1152 implementation must only remap if the requested file ranges have 1153 identical contents. If REMAP_FILE_CAN_SHORTEN is set, the caller is 1154 ok with the implementation shortening the request length to 1155 satisfy alignment or EOF requirements (or any other reason). 1156 1157``fadvise`` 1158 possibly called by the fadvise64() system call. 1159 1160Note that the file operations are implemented by the specific 1161filesystem in which the inode resides. When opening a device node 1162(character or block special) most filesystems will call special 1163support routines in the VFS which will locate the required device 1164driver information. These support routines replace the filesystem file 1165operations with those for the device driver, and then proceed to call 1166the new open() method for the file. This is how opening a device file 1167in the filesystem eventually ends up calling the device driver open() 1168method. 1169 1170 1171Directory Entry Cache (dcache) 1172============================== 1173 1174 1175struct dentry_operations 1176------------------------ 1177 1178This describes how a filesystem can overload the standard dentry 1179operations. Dentries and the dcache are the domain of the VFS and the 1180individual filesystem implementations. Device drivers have no business 1181here. These methods may be set to NULL, as they are either optional or 1182the VFS uses a default. As of kernel 2.6.22, the following members are 1183defined: 1184 1185.. code-block:: c 1186 1187 struct dentry_operations { 1188 int (*d_revalidate)(struct dentry *, unsigned int); 1189 int (*d_weak_revalidate)(struct dentry *, unsigned int); 1190 int (*d_hash)(const struct dentry *, struct qstr *); 1191 int (*d_compare)(const struct dentry *, 1192 unsigned int, const char *, const struct qstr *); 1193 int (*d_delete)(const struct dentry *); 1194 int (*d_init)(struct dentry *); 1195 void (*d_release)(struct dentry *); 1196 void (*d_iput)(struct dentry *, struct inode *); 1197 char *(*d_dname)(struct dentry *, char *, int); 1198 struct vfsmount *(*d_automount)(struct path *); 1199 int (*d_manage)(const struct path *, bool); 1200 struct dentry *(*d_real)(struct dentry *, const struct inode *); 1201 }; 1202 1203``d_revalidate`` 1204 called when the VFS needs to revalidate a dentry. This is 1205 called whenever a name look-up finds a dentry in the dcache. 1206 Most local filesystems leave this as NULL, because all their 1207 dentries in the dcache are valid. Network filesystems are 1208 different since things can change on the server without the 1209 client necessarily being aware of it. 1210 1211 This function should return a positive value if the dentry is 1212 still valid, and zero or a negative error code if it isn't. 1213 1214 d_revalidate may be called in rcu-walk mode (flags & 1215 LOOKUP_RCU). If in rcu-walk mode, the filesystem must 1216 revalidate the dentry without blocking or storing to the dentry, 1217 d_parent and d_inode should not be used without care (because 1218 they can change and, in d_inode case, even become NULL under 1219 us). 1220 1221 If a situation is encountered that rcu-walk cannot handle, 1222 return 1223 -ECHILD and it will be called again in ref-walk mode. 1224 1225``_weak_revalidate`` 1226 called when the VFS needs to revalidate a "jumped" dentry. This 1227 is called when a path-walk ends at dentry that was not acquired 1228 by doing a lookup in the parent directory. This includes "/", 1229 "." and "..", as well as procfs-style symlinks and mountpoint 1230 traversal. 1231 1232 In this case, we are less concerned with whether the dentry is 1233 still fully correct, but rather that the inode is still valid. 1234 As with d_revalidate, most local filesystems will set this to 1235 NULL since their dcache entries are always valid. 1236 1237 This function has the same return code semantics as 1238 d_revalidate. 1239 1240 d_weak_revalidate is only called after leaving rcu-walk mode. 1241 1242``d_hash`` 1243 called when the VFS adds a dentry to the hash table. The first 1244 dentry passed to d_hash is the parent directory that the name is 1245 to be hashed into. 1246 1247 Same locking and synchronisation rules as d_compare regarding 1248 what is safe to dereference etc. 1249 1250``d_compare`` 1251 called to compare a dentry name with a given name. The first 1252 dentry is the parent of the dentry to be compared, the second is 1253 the child dentry. len and name string are properties of the 1254 dentry to be compared. qstr is the name to compare it with. 1255 1256 Must be constant and idempotent, and should not take locks if 1257 possible, and should not or store into the dentry. Should not 1258 dereference pointers outside the dentry without lots of care 1259 (eg. d_parent, d_inode, d_name should not be used). 1260 1261 However, our vfsmount is pinned, and RCU held, so the dentries 1262 and inodes won't disappear, neither will our sb or filesystem 1263 module. ->d_sb may be used. 1264 1265 It is a tricky calling convention because it needs to be called 1266 under "rcu-walk", ie. without any locks or references on things. 1267 1268``d_delete`` 1269 called when the last reference to a dentry is dropped and the 1270 dcache is deciding whether or not to cache it. Return 1 to 1271 delete immediately, or 0 to cache the dentry. Default is NULL 1272 which means to always cache a reachable dentry. d_delete must 1273 be constant and idempotent. 1274 1275``d_init`` 1276 called when a dentry is allocated 1277 1278``d_release`` 1279 called when a dentry is really deallocated 1280 1281``d_iput`` 1282 called when a dentry loses its inode (just prior to its being 1283 deallocated). The default when this is NULL is that the VFS 1284 calls iput(). If you define this method, you must call iput() 1285 yourself 1286 1287``d_dname`` 1288 called when the pathname of a dentry should be generated. 1289 Useful for some pseudo filesystems (sockfs, pipefs, ...) to 1290 delay pathname generation. (Instead of doing it when dentry is 1291 created, it's done only when the path is needed.). Real 1292 filesystems probably dont want to use it, because their dentries 1293 are present in global dcache hash, so their hash should be an 1294 invariant. As no lock is held, d_dname() should not try to 1295 modify the dentry itself, unless appropriate SMP safety is used. 1296 CAUTION : d_path() logic is quite tricky. The correct way to 1297 return for example "Hello" is to put it at the end of the 1298 buffer, and returns a pointer to the first char. 1299 dynamic_dname() helper function is provided to take care of 1300 this. 1301 1302 Example : 1303 1304.. code-block:: c 1305 1306 static char *pipefs_dname(struct dentry *dent, char *buffer, int buflen) 1307 { 1308 return dynamic_dname(dentry, buffer, buflen, "pipe:[%lu]", 1309 dentry->d_inode->i_ino); 1310 } 1311 1312``d_automount`` 1313 called when an automount dentry is to be traversed (optional). 1314 This should create a new VFS mount record and return the record 1315 to the caller. The caller is supplied with a path parameter 1316 giving the automount directory to describe the automount target 1317 and the parent VFS mount record to provide inheritable mount 1318 parameters. NULL should be returned if someone else managed to 1319 make the automount first. If the vfsmount creation failed, then 1320 an error code should be returned. If -EISDIR is returned, then 1321 the directory will be treated as an ordinary directory and 1322 returned to pathwalk to continue walking. 1323 1324 If a vfsmount is returned, the caller will attempt to mount it 1325 on the mountpoint and will remove the vfsmount from its 1326 expiration list in the case of failure. The vfsmount should be 1327 returned with 2 refs on it to prevent automatic expiration - the 1328 caller will clean up the additional ref. 1329 1330 This function is only used if DCACHE_NEED_AUTOMOUNT is set on 1331 the dentry. This is set by __d_instantiate() if S_AUTOMOUNT is 1332 set on the inode being added. 1333 1334``d_manage`` 1335 called to allow the filesystem to manage the transition from a 1336 dentry (optional). This allows autofs, for example, to hold up 1337 clients waiting to explore behind a 'mountpoint' while letting 1338 the daemon go past and construct the subtree there. 0 should be 1339 returned to let the calling process continue. -EISDIR can be 1340 returned to tell pathwalk to use this directory as an ordinary 1341 directory and to ignore anything mounted on it and not to check 1342 the automount flag. Any other error code will abort pathwalk 1343 completely. 1344 1345 If the 'rcu_walk' parameter is true, then the caller is doing a 1346 pathwalk in RCU-walk mode. Sleeping is not permitted in this 1347 mode, and the caller can be asked to leave it and call again by 1348 returning -ECHILD. -EISDIR may also be returned to tell 1349 pathwalk to ignore d_automount or any mounts. 1350 1351 This function is only used if DCACHE_MANAGE_TRANSIT is set on 1352 the dentry being transited from. 1353 1354``d_real`` 1355 overlay/union type filesystems implement this method to return 1356 one of the underlying dentries hidden by the overlay. It is 1357 used in two different modes: 1358 1359 Called from file_dentry() it returns the real dentry matching 1360 the inode argument. The real dentry may be from a lower layer 1361 already copied up, but still referenced from the file. This 1362 mode is selected with a non-NULL inode argument. 1363 1364 With NULL inode the topmost real underlying dentry is returned. 1365 1366Each dentry has a pointer to its parent dentry, as well as a hash list 1367of child dentries. Child dentries are basically like files in a 1368directory. 1369 1370 1371Directory Entry Cache API 1372-------------------------- 1373 1374There are a number of functions defined which permit a filesystem to 1375manipulate dentries: 1376 1377``dget`` 1378 open a new handle for an existing dentry (this just increments 1379 the usage count) 1380 1381``dput`` 1382 close a handle for a dentry (decrements the usage count). If 1383 the usage count drops to 0, and the dentry is still in its 1384 parent's hash, the "d_delete" method is called to check whether 1385 it should be cached. If it should not be cached, or if the 1386 dentry is not hashed, it is deleted. Otherwise cached dentries 1387 are put into an LRU list to be reclaimed on memory shortage. 1388 1389``d_drop`` 1390 this unhashes a dentry from its parents hash list. A subsequent 1391 call to dput() will deallocate the dentry if its usage count 1392 drops to 0 1393 1394``d_delete`` 1395 delete a dentry. If there are no other open references to the 1396 dentry then the dentry is turned into a negative dentry (the 1397 d_iput() method is called). If there are other references, then 1398 d_drop() is called instead 1399 1400``d_add`` 1401 add a dentry to its parents hash list and then calls 1402 d_instantiate() 1403 1404``d_instantiate`` 1405 add a dentry to the alias hash list for the inode and updates 1406 the "d_inode" member. The "i_count" member in the inode 1407 structure should be set/incremented. If the inode pointer is 1408 NULL, the dentry is called a "negative dentry". This function 1409 is commonly called when an inode is created for an existing 1410 negative dentry 1411 1412``d_lookup`` 1413 look up a dentry given its parent and path name component It 1414 looks up the child of that given name from the dcache hash 1415 table. If it is found, the reference count is incremented and 1416 the dentry is returned. The caller must use dput() to free the 1417 dentry when it finishes using it. 1418 1419 1420Mount Options 1421============= 1422 1423 1424Parsing options 1425--------------- 1426 1427On mount and remount the filesystem is passed a string containing a 1428comma separated list of mount options. The options can have either of 1429these forms: 1430 1431 option 1432 option=value 1433 1434The <linux/parser.h> header defines an API that helps parse these 1435options. There are plenty of examples on how to use it in existing 1436filesystems. 1437 1438 1439Showing options 1440--------------- 1441 1442If a filesystem accepts mount options, it must define show_options() to 1443show all the currently active options. The rules are: 1444 1445 - options MUST be shown which are not default or their values differ 1446 from the default 1447 1448 - options MAY be shown which are enabled by default or have their 1449 default value 1450 1451Options used only internally between a mount helper and the kernel (such 1452as file descriptors), or which only have an effect during the mounting 1453(such as ones controlling the creation of a journal) are exempt from the 1454above rules. 1455 1456The underlying reason for the above rules is to make sure, that a mount 1457can be accurately replicated (e.g. umounting and mounting again) based 1458on the information found in /proc/mounts. 1459 1460 1461Resources 1462========= 1463 1464(Note some of these resources are not up-to-date with the latest kernel 1465 version.) 1466 1467Creating Linux virtual filesystems. 2002 1468 <https://lwn.net/Articles/13325/> 1469 1470The Linux Virtual File-system Layer by Neil Brown. 1999 1471 <http://www.cse.unsw.edu.au/~neilb/oss/linux-commentary/vfs.html> 1472 1473A tour of the Linux VFS by Michael K. Johnson. 1996 1474 <https://www.tldp.org/LDP/khg/HyperNews/get/fs/vfstour.html> 1475 1476A small trail through the Linux kernel by Andries Brouwer. 2001 1477 <https://www.win.tue.nl/~aeb/linux/vfs/trail.html> 1478