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 *, int);
13 int (*d_hash) (struct dentry *, struct qstr *);
14 int (*d_compare) (struct dentry *, struct qstr *, struct qstr *);
15 int (*d_delete)(struct dentry *);
16 void (*d_release)(struct dentry *);
17 void (*d_iput)(struct dentry *, struct inode *);
18 char *(*d_dname)((struct dentry *dentry, char *buffer, int buflen);
21 dcache_lock rename_lock ->d_lock may block
22 d_revalidate: no no no yes
24 d_compare: no yes no no
25 d_delete: yes no yes no
26 d_release: no no no yes
30 --------------------------- inode_operations ---------------------------
32 int (*create) (struct inode *,struct dentry *,int, struct nameidata *);
33 struct dentry * (*lookup) (struct inode *,struct dentry *, struct nameid
35 int (*link) (struct dentry *,struct inode *,struct dentry *);
36 int (*unlink) (struct inode *,struct dentry *);
37 int (*symlink) (struct inode *,struct dentry *,const char *);
38 int (*mkdir) (struct inode *,struct dentry *,int);
39 int (*rmdir) (struct inode *,struct dentry *);
40 int (*mknod) (struct inode *,struct dentry *,int,dev_t);
41 int (*rename) (struct inode *, struct dentry *,
42 struct inode *, struct dentry *);
43 int (*readlink) (struct dentry *, char __user *,int);
44 void * (*follow_link) (struct dentry *, struct nameidata *);
45 void (*put_link) (struct dentry *, struct nameidata *, void *);
46 void (*truncate) (struct inode *);
47 int (*permission) (struct inode *, int, struct nameidata *);
48 int (*check_acl)(struct inode *, int);
49 int (*setattr) (struct dentry *, struct iattr *);
50 int (*getattr) (struct vfsmount *, struct dentry *, struct kstat *);
51 int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
52 ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t);
53 ssize_t (*listxattr) (struct dentry *, char *, size_t);
54 int (*removexattr) (struct dentry *, const char *);
55 void (*truncate_range)(struct inode *, loff_t, loff_t);
56 long (*fallocate)(struct inode *inode, int mode, loff_t offset, loff_t len);
57 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start, u64 len);
69 rmdir: yes (both) (see below)
70 rename: yes (all) (see below)
74 truncate: yes (see below)
86 Additionally, ->rmdir(), ->unlink() and ->rename() have ->i_mutex on
88 cross-directory ->rename() has (per-superblock) ->s_vfs_rename_sem.
89 ->truncate() is never called directly - it's a callback, not a
90 method. It's called by vmtruncate() - deprecated library function used by
91 ->setattr(). Locking information above applies to that call (i.e. is
92 inherited from ->setattr() - vmtruncate() is used when ATTR_SIZE had been
95 See Documentation/filesystems/directory-locking for more detailed discussion
96 of the locking scheme for directory operations.
98 --------------------------- super_operations ---------------------------
100 struct inode *(*alloc_inode)(struct super_block *sb);
101 void (*destroy_inode)(struct inode *);
102 void (*dirty_inode) (struct inode *);
103 int (*write_inode) (struct inode *, struct writeback_control *wbc);
104 int (*drop_inode) (struct inode *);
105 void (*evict_inode) (struct inode *);
106 void (*put_super) (struct super_block *);
107 void (*write_super) (struct super_block *);
108 int (*sync_fs)(struct super_block *sb, int wait);
109 int (*freeze_fs) (struct super_block *);
110 int (*unfreeze_fs) (struct super_block *);
111 int (*statfs) (struct dentry *, struct kstatfs *);
112 int (*remount_fs) (struct super_block *, int *, char *);
113 void (*umount_begin) (struct super_block *);
114 int (*show_options)(struct seq_file *, struct vfsmount *);
115 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
116 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
117 int (*bdev_try_to_free_page)(struct super_block*, struct page*, gfp_t);
120 All may block [not true, see below]
124 dirty_inode: (must not sleep)
126 drop_inode: !!!inode_lock!!!
133 statfs: maybe(read) (see below)
136 show_options: no (namespace_sem)
137 quota_read: no (see below)
138 quota_write: no (see below)
139 bdev_try_to_free_page: no (see below)
141 ->statfs() has s_umount (shared) when called by ustat(2) (native or
142 compat), but that's an accident of bad API; s_umount is used to pin
143 the superblock down when we only have dev_t given us by userland to
144 identify the superblock. Everything else (statfs(), fstatfs(), etc.)
145 doesn't hold it when calling ->statfs() - superblock is pinned down
146 by resolving the pathname passed to syscall.
147 ->quota_read() and ->quota_write() functions are both guaranteed to
148 be the only ones operating on the quota file by the quota code (via
149 dqio_sem) (unless an admin really wants to screw up something and
150 writes to quota files with quotas on). For other details about locking
151 see also dquot_operations section.
152 ->bdev_try_to_free_page is called from the ->releasepage handler of
153 the block device inode. See there for more details.
155 --------------------------- file_system_type ---------------------------
157 int (*get_sb) (struct file_system_type *, int,
158 const char *, void *, struct vfsmount *);
159 struct dentry *(*mount) (struct file_system_type *, int,
160 const char *, void *);
161 void (*kill_sb) (struct super_block *);
168 ->get_sb() returns error or 0 with locked superblock attached to the vfsmount
169 (exclusive on ->s_umount).
170 ->mount() returns ERR_PTR or the root dentry.
171 ->kill_sb() takes a write-locked superblock, does all shutdown work on it,
172 unlocks and drops the reference.
174 --------------------------- address_space_operations --------------------------
176 int (*writepage)(struct page *page, struct writeback_control *wbc);
177 int (*readpage)(struct file *, struct page *);
178 int (*sync_page)(struct page *);
179 int (*writepages)(struct address_space *, struct writeback_control *);
180 int (*set_page_dirty)(struct page *page);
181 int (*readpages)(struct file *filp, struct address_space *mapping,
182 struct list_head *pages, unsigned nr_pages);
183 int (*write_begin)(struct file *, struct address_space *mapping,
184 loff_t pos, unsigned len, unsigned flags,
185 struct page **pagep, void **fsdata);
186 int (*write_end)(struct file *, struct address_space *mapping,
187 loff_t pos, unsigned len, unsigned copied,
188 struct page *page, void *fsdata);
189 sector_t (*bmap)(struct address_space *, sector_t);
190 int (*invalidatepage) (struct page *, unsigned long);
191 int (*releasepage) (struct page *, int);
192 void (*freepage)(struct page *);
193 int (*direct_IO)(int, struct kiocb *, const struct iovec *iov,
194 loff_t offset, unsigned long nr_segs);
195 int (*get_xip_mem)(struct address_space *, pgoff_t, int, void **,
197 int (*migratepage)(struct address_space *, struct page *, struct page *);
198 int (*launder_page)(struct page *);
199 int (*is_partially_uptodate)(struct page *, read_descriptor_t *, unsigned long);
200 int (*error_remove_page)(struct address_space *, struct page *);
203 All except set_page_dirty and freepage may block
205 PageLocked(page) i_mutex
206 writepage: yes, unlocks (see below)
207 readpage: yes, unlocks
212 write_begin: locks the page yes
213 write_end: yes, unlocks yes
220 migratepage: yes (both)
222 is_partially_uptodate: yes
223 error_remove_page: yes
225 ->write_begin(), ->write_end(), ->sync_page() and ->readpage()
226 may be called from the request handler (/dev/loop).
228 ->readpage() unlocks the page, either synchronously or via I/O
231 ->readpages() populates the pagecache with the passed pages and starts
232 I/O against them. They come unlocked upon I/O completion.
234 ->writepage() is used for two purposes: for "memory cleansing" and for
235 "sync". These are quite different operations and the behaviour may differ
236 depending upon the mode.
238 If writepage is called for sync (wbc->sync_mode != WBC_SYNC_NONE) then
239 it *must* start I/O against the page, even if that would involve
240 blocking on in-progress I/O.
242 If writepage is called for memory cleansing (sync_mode ==
243 WBC_SYNC_NONE) then its role is to get as much writeout underway as
244 possible. So writepage should try to avoid blocking against
245 currently-in-progress I/O.
247 If the filesystem is not called for "sync" and it determines that it
248 would need to block against in-progress I/O to be able to start new I/O
249 against the page the filesystem should redirty the page with
250 redirty_page_for_writepage(), then unlock the page and return zero.
251 This may also be done to avoid internal deadlocks, but rarely.
253 If the filesystem is called for sync then it must wait on any
254 in-progress I/O and then start new I/O.
256 The filesystem should unlock the page synchronously, before returning to the
257 caller, unless ->writepage() returns special WRITEPAGE_ACTIVATE
258 value. WRITEPAGE_ACTIVATE means that page cannot really be written out
259 currently, and VM should stop calling ->writepage() on this page for some
260 time. VM does this by moving page to the head of the active list, hence the
263 Unless the filesystem is going to redirty_page_for_writepage(), unlock the page
264 and return zero, writepage *must* run set_page_writeback() against the page,
265 followed by unlocking it. Once set_page_writeback() has been run against the
266 page, write I/O can be submitted and the write I/O completion handler must run
267 end_page_writeback() once the I/O is complete. If no I/O is submitted, the
268 filesystem must run end_page_writeback() against the page before returning from
271 That is: after 2.5.12, pages which are under writeout are *not* locked. Note,
272 if the filesystem needs the page to be locked during writeout, that is ok, too,
273 the page is allowed to be unlocked at any point in time between the calls to
274 set_page_writeback() and end_page_writeback().
276 Note, failure to run either redirty_page_for_writepage() or the combination of
277 set_page_writeback()/end_page_writeback() on a page submitted to writepage
278 will leave the page itself marked clean but it will be tagged as dirty in the
279 radix tree. This incoherency can lead to all sorts of hard-to-debug problems
280 in the filesystem like having dirty inodes at umount and losing written data.
282 ->sync_page() locking rules are not well-defined - usually it is called
283 with lock on page, but that is not guaranteed. Considering the currently
284 existing instances of this method ->sync_page() itself doesn't look
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 ----------------------- file_lock_operations ------------------------------
327 void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
328 void (*fl_release_private)(struct file_lock *);
332 file_lock_lock may block
334 fl_release_private: maybe no
336 ----------------------- lock_manager_operations ---------------------------
338 int (*fl_compare_owner)(struct file_lock *, struct file_lock *);
339 void (*fl_notify)(struct file_lock *); /* unblock callback */
340 int (*fl_grant)(struct file_lock *, struct file_lock *, int);
341 void (*fl_release_private)(struct file_lock *);
342 void (*fl_break)(struct file_lock *); /* break_lease callback */
343 int (*fl_mylease)(struct file_lock *, struct file_lock *);
344 int (*fl_change)(struct file_lock **, int);
347 file_lock_lock may block
348 fl_compare_owner: yes no
351 fl_release_private: maybe no
356 --------------------------- buffer_head -----------------------------------
358 void (*b_end_io)(struct buffer_head *bh, int uptodate);
361 called from interrupts. In other words, extreme care is needed here.
362 bh is locked, but that's all warranties we have here. Currently only RAID1,
363 highmem, fs/buffer.c, and fs/ntfs/aops.c are providing these. Block devices
364 call this method upon the IO completion.
366 --------------------------- block_device_operations -----------------------
368 int (*open) (struct block_device *, fmode_t);
369 int (*release) (struct gendisk *, fmode_t);
370 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
371 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
372 int (*direct_access) (struct block_device *, sector_t, void **, unsigned long *);
373 int (*media_changed) (struct gendisk *);
374 void (*unlock_native_capacity) (struct gendisk *);
375 int (*revalidate_disk) (struct gendisk *);
376 int (*getgeo)(struct block_device *, struct hd_geometry *);
377 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
387 unlock_native_capacity: no
390 swap_slot_free_notify: no (see below)
392 media_changed, unlock_native_capacity and revalidate_disk are called only from
395 swap_slot_free_notify is called with swap_lock and sometimes the page lock
399 --------------------------- file_operations -------------------------------
401 loff_t (*llseek) (struct file *, loff_t, int);
402 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
403 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
404 ssize_t (*aio_read) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
405 ssize_t (*aio_write) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
406 int (*readdir) (struct file *, void *, filldir_t);
407 unsigned int (*poll) (struct file *, struct poll_table_struct *);
408 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
409 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
410 int (*mmap) (struct file *, struct vm_area_struct *);
411 int (*open) (struct inode *, struct file *);
412 int (*flush) (struct file *);
413 int (*release) (struct inode *, struct file *);
414 int (*fsync) (struct file *, int datasync);
415 int (*aio_fsync) (struct kiocb *, int datasync);
416 int (*fasync) (int, struct file *, int);
417 int (*lock) (struct file *, int, struct file_lock *);
418 ssize_t (*readv) (struct file *, const struct iovec *, unsigned long,
420 ssize_t (*writev) (struct file *, const struct iovec *, unsigned long,
422 ssize_t (*sendfile) (struct file *, loff_t *, size_t, read_actor_t,
424 ssize_t (*sendpage) (struct file *, struct page *, int, size_t,
426 unsigned long (*get_unmapped_area)(struct file *, unsigned long,
427 unsigned long, unsigned long, unsigned long);
428 int (*check_flags)(int);
429 int (*flock) (struct file *, int, struct file_lock *);
430 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *,
431 size_t, unsigned int);
432 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *,
433 size_t, unsigned int);
434 int (*setlease)(struct file *, long, struct file_lock **);
438 All may block except for ->setlease.
439 No VFS locks held on entry except for ->fsync and ->setlease.
441 ->fsync() has i_mutex on inode.
443 ->setlease has the file_list_lock held and must not sleep.
445 ->llseek() locking has moved from llseek to the individual llseek
446 implementations. If your fs is not using generic_file_llseek, you
447 need to acquire and release the appropriate locks in your ->llseek().
448 For many filesystems, it is probably safe to acquire the inode
449 mutex or just to use i_size_read() instead.
450 Note: this does not protect the file->f_pos against concurrent modifications
451 since this is something the userspace has to take care about.
453 ->fasync() is responsible for maintaining the FASYNC bit in filp->f_flags.
454 Most instances call fasync_helper(), which does that maintenance, so it's
455 not normally something one needs to worry about. Return values > 0 will be
456 mapped to zero in the VFS layer.
458 ->readdir() and ->ioctl() on directories must be changed. Ideally we would
459 move ->readdir() to inode_operations and use a separate method for directory
460 ->ioctl() or kill the latter completely. One of the problems is that for
461 anything that resembles union-mount we won't have a struct file for all
462 components. And there are other reasons why the current interface is a mess...
464 ->read on directories probably must go away - we should just enforce -EISDIR
465 in sys_read() and friends.
467 --------------------------- dquot_operations -------------------------------
469 int (*write_dquot) (struct dquot *);
470 int (*acquire_dquot) (struct dquot *);
471 int (*release_dquot) (struct dquot *);
472 int (*mark_dirty) (struct dquot *);
473 int (*write_info) (struct super_block *, int);
475 These operations are intended to be more or less wrapping functions that ensure
476 a proper locking wrt the filesystem and call the generic quota operations.
478 What filesystem should expect from the generic quota functions:
480 FS recursion Held locks when called
481 write_dquot: yes dqonoff_sem or dqptr_sem
482 acquire_dquot: yes dqonoff_sem or dqptr_sem
483 release_dquot: yes dqonoff_sem or dqptr_sem
485 write_info: yes dqonoff_sem
487 FS recursion means calling ->quota_read() and ->quota_write() from superblock
490 More details about quota locking can be found in fs/dquot.c.
492 --------------------------- vm_operations_struct -----------------------------
494 void (*open)(struct vm_area_struct*);
495 void (*close)(struct vm_area_struct*);
496 int (*fault)(struct vm_area_struct*, struct vm_fault *);
497 int (*page_mkwrite)(struct vm_area_struct *, struct vm_fault *);
498 int (*access)(struct vm_area_struct *, unsigned long, void*, int, int);
501 mmap_sem PageLocked(page)
504 fault: yes can return with page locked
505 page_mkwrite: yes can return with page locked
508 ->fault() is called when a previously not present pte is about
509 to be faulted in. The filesystem must find and return the page associated
510 with the passed in "pgoff" in the vm_fault structure. If it is possible that
511 the page may be truncated and/or invalidated, then the filesystem must lock
512 the page, then ensure it is not already truncated (the page lock will block
513 subsequent truncate), and then return with VM_FAULT_LOCKED, and the page
514 locked. The VM will unlock the page.
516 ->page_mkwrite() is called when a previously read-only pte is
517 about to become writeable. The filesystem again must ensure that there are
518 no truncate/invalidate races, and then return with the page locked. If
519 the page has been truncated, the filesystem should not look up a new page
520 like the ->fault() handler, but simply return with VM_FAULT_NOPAGE, which
521 will cause the VM to retry the fault.
523 ->access() is called when get_user_pages() fails in
524 acces_process_vm(), typically used to debug a process through
525 /proc/pid/mem or ptrace. This function is needed only for
526 VM_IO | VM_PFNMAP VMAs.
528 ================================================================================
531 (if you break something or notice that it is broken and do not fix it yourself
532 - at least put it here)