1 The text below describes the locking rules for VFS-related methods.
2 It is (believed to be) up-to-date. *Please*, if you change anything in
3 prototypes or locking protocols - update this file. And update the relevant
4 instances in the tree, don't leave that to maintainers of filesystems/devices/
5 etc. At the very least, put the list of dubious cases in the end of this file.
6 Don't turn it into log - maintainers of out-of-the-tree code are supposed to
7 be able to use diff(1).
8 Thing currently missing here: socket operations. Alexey?
10 --------------------------- dentry_operations --------------------------
12 int (*d_revalidate)(struct dentry *, unsigned int);
13 int (*d_weak_revalidate)(struct dentry *, unsigned int);
14 int (*d_hash)(const struct dentry *, struct qstr *);
15 int (*d_compare)(const struct dentry *, const struct dentry *,
16 unsigned int, const char *, const struct qstr *);
17 int (*d_delete)(struct dentry *);
18 int (*d_init)(struct dentry *);
19 void (*d_release)(struct dentry *);
20 void (*d_iput)(struct dentry *, struct inode *);
21 char *(*d_dname)((struct dentry *dentry, char *buffer, int buflen);
22 struct vfsmount *(*d_automount)(struct path *path);
23 int (*d_manage)(struct dentry *, bool);
24 struct dentry *(*d_real)(struct dentry *, const struct inode *,
28 rename_lock ->d_lock may block rcu-walk
29 d_revalidate: no no yes (ref-walk) maybe
30 d_weak_revalidate:no no yes no
32 d_compare: yes no no maybe
33 d_delete: no yes no no
35 d_release: no no yes no
39 d_automount: no no yes no
40 d_manage: no no yes (ref-walk) maybe
43 --------------------------- inode_operations ---------------------------
45 int (*create) (struct inode *,struct dentry *,umode_t, bool);
46 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
47 int (*link) (struct dentry *,struct inode *,struct dentry *);
48 int (*unlink) (struct inode *,struct dentry *);
49 int (*symlink) (struct inode *,struct dentry *,const char *);
50 int (*mkdir) (struct inode *,struct dentry *,umode_t);
51 int (*rmdir) (struct inode *,struct dentry *);
52 int (*mknod) (struct inode *,struct dentry *,umode_t,dev_t);
53 int (*rename) (struct inode *, struct dentry *,
54 struct inode *, struct dentry *);
55 int (*rename2) (struct inode *, struct dentry *,
56 struct inode *, struct dentry *, unsigned int);
57 int (*readlink) (struct dentry *, char __user *,int);
58 const char *(*get_link) (struct dentry *, struct inode *, void **);
59 void (*truncate) (struct inode *);
60 int (*permission) (struct inode *, int, unsigned int);
61 int (*get_acl)(struct inode *, int);
62 int (*setattr) (struct dentry *, struct iattr *);
63 int (*getattr) (struct vfsmount *, struct dentry *, struct kstat *);
64 int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
65 ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t);
66 ssize_t (*listxattr) (struct dentry *, char *, size_t);
67 int (*removexattr) (struct dentry *, const char *);
68 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start, u64 len);
69 void (*update_time)(struct inode *, struct timespec *, int);
70 int (*atomic_open)(struct inode *, struct dentry *,
71 struct file *, unsigned open_flag,
72 umode_t create_mode, int *opened);
73 int (*tmpfile) (struct inode *, struct dentry *, umode_t);
85 rmdir: yes (both) (see below)
86 rename: yes (all) (see below)
87 rename2: yes (all) (see below)
91 permission: no (may not block if called in rcu-walk mode)
103 Additionally, ->rmdir(), ->unlink() and ->rename() have ->i_mutex on
105 cross-directory ->rename() and rename2() has (per-superblock)
108 See Documentation/filesystems/directory-locking for more detailed discussion
109 of the locking scheme for directory operations.
111 --------------------------- super_operations ---------------------------
113 struct inode *(*alloc_inode)(struct super_block *sb);
114 void (*destroy_inode)(struct inode *);
115 void (*dirty_inode) (struct inode *, int flags);
116 int (*write_inode) (struct inode *, struct writeback_control *wbc);
117 int (*drop_inode) (struct inode *);
118 void (*evict_inode) (struct inode *);
119 void (*put_super) (struct super_block *);
120 int (*sync_fs)(struct super_block *sb, int wait);
121 int (*freeze_fs) (struct super_block *);
122 int (*unfreeze_fs) (struct super_block *);
123 int (*statfs) (struct dentry *, struct kstatfs *);
124 int (*remount_fs) (struct super_block *, int *, char *);
125 void (*umount_begin) (struct super_block *);
126 int (*show_options)(struct seq_file *, struct dentry *);
127 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
128 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
129 int (*bdev_try_to_free_page)(struct super_block*, struct page*, gfp_t);
132 All may block [not true, see below]
138 drop_inode: !!!inode->i_lock!!!
144 statfs: maybe(read) (see below)
147 show_options: no (namespace_sem)
148 quota_read: no (see below)
149 quota_write: no (see below)
150 bdev_try_to_free_page: no (see below)
152 ->statfs() has s_umount (shared) when called by ustat(2) (native or
153 compat), but that's an accident of bad API; s_umount is used to pin
154 the superblock down when we only have dev_t given us by userland to
155 identify the superblock. Everything else (statfs(), fstatfs(), etc.)
156 doesn't hold it when calling ->statfs() - superblock is pinned down
157 by resolving the pathname passed to syscall.
158 ->quota_read() and ->quota_write() functions are both guaranteed to
159 be the only ones operating on the quota file by the quota code (via
160 dqio_sem) (unless an admin really wants to screw up something and
161 writes to quota files with quotas on). For other details about locking
162 see also dquot_operations section.
163 ->bdev_try_to_free_page is called from the ->releasepage handler of
164 the block device inode. See there for more details.
166 --------------------------- file_system_type ---------------------------
168 struct dentry *(*mount) (struct file_system_type *, int,
169 const char *, void *);
170 void (*kill_sb) (struct super_block *);
176 ->mount() returns ERR_PTR or the root dentry; its superblock should be locked
178 ->kill_sb() takes a write-locked superblock, does all shutdown work on it,
179 unlocks and drops the reference.
181 --------------------------- address_space_operations --------------------------
183 int (*writepage)(struct page *page, struct writeback_control *wbc);
184 int (*readpage)(struct file *, struct page *);
185 int (*writepages)(struct address_space *, struct writeback_control *);
186 int (*set_page_dirty)(struct page *page);
187 int (*readpages)(struct file *filp, struct address_space *mapping,
188 struct list_head *pages, unsigned nr_pages);
189 int (*write_begin)(struct file *, struct address_space *mapping,
190 loff_t pos, unsigned len, unsigned flags,
191 struct page **pagep, void **fsdata);
192 int (*write_end)(struct file *, struct address_space *mapping,
193 loff_t pos, unsigned len, unsigned copied,
194 struct page *page, void *fsdata);
195 sector_t (*bmap)(struct address_space *, sector_t);
196 void (*invalidatepage) (struct page *, unsigned int, unsigned int);
197 int (*releasepage) (struct page *, int);
198 void (*freepage)(struct page *);
199 int (*direct_IO)(struct kiocb *, struct iov_iter *iter);
200 int (*migratepage)(struct address_space *, struct page *, struct page *);
201 int (*launder_page)(struct page *);
202 int (*is_partially_uptodate)(struct page *, unsigned long, unsigned long);
203 int (*error_remove_page)(struct address_space *, struct page *);
204 int (*swap_activate)(struct file *);
205 int (*swap_deactivate)(struct file *);
208 All except set_page_dirty and freepage may block
210 PageLocked(page) i_mutex
211 writepage: yes, unlocks (see below)
212 readpage: yes, unlocks
216 write_begin: locks the page yes
217 write_end: yes, unlocks yes
223 migratepage: yes (both)
225 is_partially_uptodate: yes
226 error_remove_page: yes
230 ->write_begin(), ->write_end() and ->readpage() may be called from
231 the request handler (/dev/loop).
233 ->readpage() unlocks the page, either synchronously or via I/O
236 ->readpages() populates the pagecache with the passed pages and starts
237 I/O against them. They come unlocked upon I/O completion.
239 ->writepage() is used for two purposes: for "memory cleansing" and for
240 "sync". These are quite different operations and the behaviour may differ
241 depending upon the mode.
243 If writepage is called for sync (wbc->sync_mode != WBC_SYNC_NONE) then
244 it *must* start I/O against the page, even if that would involve
245 blocking on in-progress I/O.
247 If writepage is called for memory cleansing (sync_mode ==
248 WBC_SYNC_NONE) then its role is to get as much writeout underway as
249 possible. So writepage should try to avoid blocking against
250 currently-in-progress I/O.
252 If the filesystem is not called for "sync" and it determines that it
253 would need to block against in-progress I/O to be able to start new I/O
254 against the page the filesystem should redirty the page with
255 redirty_page_for_writepage(), then unlock the page and return zero.
256 This may also be done to avoid internal deadlocks, but rarely.
258 If the filesystem is called for sync then it must wait on any
259 in-progress I/O and then start new I/O.
261 The filesystem should unlock the page synchronously, before returning to the
262 caller, unless ->writepage() returns special WRITEPAGE_ACTIVATE
263 value. WRITEPAGE_ACTIVATE means that page cannot really be written out
264 currently, and VM should stop calling ->writepage() on this page for some
265 time. VM does this by moving page to the head of the active list, hence the
268 Unless the filesystem is going to redirty_page_for_writepage(), unlock the page
269 and return zero, writepage *must* run set_page_writeback() against the page,
270 followed by unlocking it. Once set_page_writeback() has been run against the
271 page, write I/O can be submitted and the write I/O completion handler must run
272 end_page_writeback() once the I/O is complete. If no I/O is submitted, the
273 filesystem must run end_page_writeback() against the page before returning from
276 That is: after 2.5.12, pages which are under writeout are *not* locked. Note,
277 if the filesystem needs the page to be locked during writeout, that is ok, too,
278 the page is allowed to be unlocked at any point in time between the calls to
279 set_page_writeback() and end_page_writeback().
281 Note, failure to run either redirty_page_for_writepage() or the combination of
282 set_page_writeback()/end_page_writeback() on a page submitted to writepage
283 will leave the page itself marked clean but it will be tagged as dirty in the
284 radix tree. This incoherency can lead to all sorts of hard-to-debug problems
285 in the filesystem like having dirty inodes at umount and losing written data.
287 ->writepages() is used for periodic writeback and for syscall-initiated
288 sync operations. The address_space should start I/O against at least
289 *nr_to_write pages. *nr_to_write must be decremented for each page which is
290 written. The address_space implementation may write more (or less) pages
291 than *nr_to_write asks for, but it should try to be reasonably close. If
292 nr_to_write is NULL, all dirty pages must be written.
294 writepages should _only_ write pages which are present on
297 ->set_page_dirty() is called from various places in the kernel
298 when the target page is marked as needing writeback. It may be called
299 under spinlock (it cannot block) and is sometimes called with the page
302 ->bmap() is currently used by legacy ioctl() (FIBMAP) provided by some
303 filesystems and by the swapper. The latter will eventually go away. Please,
304 keep it that way and don't breed new callers.
306 ->invalidatepage() is called when the filesystem must attempt to drop
307 some or all of the buffers from the page when it is being truncated. It
308 returns zero on success. If ->invalidatepage is zero, the kernel uses
309 block_invalidatepage() instead.
311 ->releasepage() is called when the kernel is about to try to drop the
312 buffers from the page in preparation for freeing it. It returns zero to
313 indicate that the buffers are (or may be) freeable. If ->releasepage is zero,
314 the kernel assumes that the fs has no private interest in the buffers.
316 ->freepage() is called when the kernel is done dropping the page
319 ->launder_page() may be called prior to releasing a page if
320 it is still found to be dirty. It returns zero if the page was successfully
321 cleaned, or an error value if not. Note that in order to prevent the page
322 getting mapped back in and redirtied, it needs to be kept locked
323 across the entire operation.
325 ->swap_activate will be called with a non-zero argument on
326 files backing (non block device backed) swapfiles. A return value
327 of zero indicates success, in which case this file can be used for
328 backing swapspace. The swapspace operations will be proxied to the
329 address space operations.
331 ->swap_deactivate() will be called in the sys_swapoff()
332 path after ->swap_activate() returned success.
334 ----------------------- file_lock_operations ------------------------------
336 void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
337 void (*fl_release_private)(struct file_lock *);
341 inode->i_lock may block
343 fl_release_private: maybe maybe[1]
345 [1]: ->fl_release_private for flock or POSIX locks is currently allowed
346 to block. Leases however can still be freed while the i_lock is held and
347 so fl_release_private called on a lease should not block.
349 ----------------------- lock_manager_operations ---------------------------
351 int (*lm_compare_owner)(struct file_lock *, struct file_lock *);
352 unsigned long (*lm_owner_key)(struct file_lock *);
353 void (*lm_notify)(struct file_lock *); /* unblock callback */
354 int (*lm_grant)(struct file_lock *, struct file_lock *, int);
355 void (*lm_break)(struct file_lock *); /* break_lease callback */
356 int (*lm_change)(struct file_lock **, int);
360 inode->i_lock blocked_lock_lock may block
361 lm_compare_owner: yes[1] maybe no
362 lm_owner_key yes[1] yes no
363 lm_notify: yes yes no
368 [1]: ->lm_compare_owner and ->lm_owner_key are generally called with
369 *an* inode->i_lock held. It may not be the i_lock of the inode
370 associated with either file_lock argument! This is the case with deadlock
371 detection, since the code has to chase down the owners of locks that may
372 be entirely unrelated to the one on which the lock is being acquired.
373 For deadlock detection however, the blocked_lock_lock is also held. The
374 fact that these locks are held ensures that the file_locks do not
375 disappear out from under you while doing the comparison or generating an
378 --------------------------- buffer_head -----------------------------------
380 void (*b_end_io)(struct buffer_head *bh, int uptodate);
383 called from interrupts. In other words, extreme care is needed here.
384 bh is locked, but that's all warranties we have here. Currently only RAID1,
385 highmem, fs/buffer.c, and fs/ntfs/aops.c are providing these. Block devices
386 call this method upon the IO completion.
388 --------------------------- block_device_operations -----------------------
390 int (*open) (struct block_device *, fmode_t);
391 int (*release) (struct gendisk *, fmode_t);
392 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
393 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
394 int (*direct_access) (struct block_device *, sector_t, void __pmem **,
396 int (*media_changed) (struct gendisk *);
397 void (*unlock_native_capacity) (struct gendisk *);
398 int (*revalidate_disk) (struct gendisk *);
399 int (*getgeo)(struct block_device *, struct hd_geometry *);
400 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
410 unlock_native_capacity: no
413 swap_slot_free_notify: no (see below)
415 media_changed, unlock_native_capacity and revalidate_disk are called only from
418 swap_slot_free_notify is called with swap_lock and sometimes the page lock
422 --------------------------- file_operations -------------------------------
424 loff_t (*llseek) (struct file *, loff_t, int);
425 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
426 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
427 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
428 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
429 int (*iterate) (struct file *, struct dir_context *);
430 unsigned int (*poll) (struct file *, struct poll_table_struct *);
431 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
432 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
433 int (*mmap) (struct file *, struct vm_area_struct *);
434 int (*open) (struct inode *, struct file *);
435 int (*flush) (struct file *);
436 int (*release) (struct inode *, struct file *);
437 int (*fsync) (struct file *, loff_t start, loff_t end, int datasync);
438 int (*aio_fsync) (struct kiocb *, int datasync);
439 int (*fasync) (int, struct file *, int);
440 int (*lock) (struct file *, int, struct file_lock *);
441 ssize_t (*readv) (struct file *, const struct iovec *, unsigned long,
443 ssize_t (*writev) (struct file *, const struct iovec *, unsigned long,
445 ssize_t (*sendfile) (struct file *, loff_t *, size_t, read_actor_t,
447 ssize_t (*sendpage) (struct file *, struct page *, int, size_t,
449 unsigned long (*get_unmapped_area)(struct file *, unsigned long,
450 unsigned long, unsigned long, unsigned long);
451 int (*check_flags)(int);
452 int (*flock) (struct file *, int, struct file_lock *);
453 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *,
454 size_t, unsigned int);
455 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *,
456 size_t, unsigned int);
457 int (*setlease)(struct file *, long, struct file_lock **, void **);
458 long (*fallocate)(struct file *, int, loff_t, loff_t);
464 ->llseek() locking has moved from llseek to the individual llseek
465 implementations. If your fs is not using generic_file_llseek, you
466 need to acquire and release the appropriate locks in your ->llseek().
467 For many filesystems, it is probably safe to acquire the inode
468 mutex or just to use i_size_read() instead.
469 Note: this does not protect the file->f_pos against concurrent modifications
470 since this is something the userspace has to take care about.
472 ->fasync() is responsible for maintaining the FASYNC bit in filp->f_flags.
473 Most instances call fasync_helper(), which does that maintenance, so it's
474 not normally something one needs to worry about. Return values > 0 will be
475 mapped to zero in the VFS layer.
477 ->readdir() and ->ioctl() on directories must be changed. Ideally we would
478 move ->readdir() to inode_operations and use a separate method for directory
479 ->ioctl() or kill the latter completely. One of the problems is that for
480 anything that resembles union-mount we won't have a struct file for all
481 components. And there are other reasons why the current interface is a mess...
483 ->read on directories probably must go away - we should just enforce -EISDIR
484 in sys_read() and friends.
486 ->setlease operations should call generic_setlease() before or after setting
487 the lease within the individual filesystem to record the result of the
490 --------------------------- dquot_operations -------------------------------
492 int (*write_dquot) (struct dquot *);
493 int (*acquire_dquot) (struct dquot *);
494 int (*release_dquot) (struct dquot *);
495 int (*mark_dirty) (struct dquot *);
496 int (*write_info) (struct super_block *, int);
498 These operations are intended to be more or less wrapping functions that ensure
499 a proper locking wrt the filesystem and call the generic quota operations.
501 What filesystem should expect from the generic quota functions:
503 FS recursion Held locks when called
504 write_dquot: yes dqonoff_sem or dqptr_sem
505 acquire_dquot: yes dqonoff_sem or dqptr_sem
506 release_dquot: yes dqonoff_sem or dqptr_sem
508 write_info: yes dqonoff_sem
510 FS recursion means calling ->quota_read() and ->quota_write() from superblock
513 More details about quota locking can be found in fs/dquot.c.
515 --------------------------- vm_operations_struct -----------------------------
517 void (*open)(struct vm_area_struct*);
518 void (*close)(struct vm_area_struct*);
519 int (*fault)(struct vm_area_struct*, struct vm_fault *);
520 int (*page_mkwrite)(struct vm_area_struct *, struct vm_fault *);
521 int (*pfn_mkwrite)(struct vm_area_struct *, struct vm_fault *);
522 int (*access)(struct vm_area_struct *, unsigned long, void*, int, int);
525 mmap_sem PageLocked(page)
528 fault: yes can return with page locked
530 page_mkwrite: yes can return with page locked
534 ->fault() is called when a previously not present pte is about
535 to be faulted in. The filesystem must find and return the page associated
536 with the passed in "pgoff" in the vm_fault structure. If it is possible that
537 the page may be truncated and/or invalidated, then the filesystem must lock
538 the page, then ensure it is not already truncated (the page lock will block
539 subsequent truncate), and then return with VM_FAULT_LOCKED, and the page
540 locked. The VM will unlock the page.
542 ->map_pages() is called when VM asks to map easy accessible pages.
543 Filesystem should find and map pages associated with offsets from "pgoff"
544 till "max_pgoff". ->map_pages() is called with page table locked and must
545 not block. If it's not possible to reach a page without blocking,
546 filesystem should skip it. Filesystem should use do_set_pte() to setup
547 page table entry. Pointer to entry associated with offset "pgoff" is
548 passed in "pte" field in vm_fault structure. Pointers to entries for other
549 offsets should be calculated relative to "pte".
551 ->page_mkwrite() is called when a previously read-only pte is
552 about to become writeable. The filesystem again must ensure that there are
553 no truncate/invalidate races, and then return with the page locked. If
554 the page has been truncated, the filesystem should not look up a new page
555 like the ->fault() handler, but simply return with VM_FAULT_NOPAGE, which
556 will cause the VM to retry the fault.
558 ->pfn_mkwrite() is the same as page_mkwrite but when the pte is
559 VM_PFNMAP or VM_MIXEDMAP with a page-less entry. Expected return is
560 VM_FAULT_NOPAGE. Or one of the VM_FAULT_ERROR types. The default behavior
561 after this call is to make the pte read-write, unless pfn_mkwrite returns
564 ->access() is called when get_user_pages() fails in
565 access_process_vm(), typically used to debug a process through
566 /proc/pid/mem or ptrace. This function is needed only for
567 VM_IO | VM_PFNMAP VMAs.
569 ================================================================================
572 (if you break something or notice that it is broken and do not fix it yourself
573 - at least put it here)