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 *);
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
29 --------------------------- inode_operations ---------------------------
31 int (*create) (struct inode *,struct dentry *,int, struct nameidata *);
32 struct dentry * (*lookup) (struct inode *,struct dentry *, struct nameid
34 int (*link) (struct dentry *,struct inode *,struct dentry *);
35 int (*unlink) (struct inode *,struct dentry *);
36 int (*symlink) (struct inode *,struct dentry *,const char *);
37 int (*mkdir) (struct inode *,struct dentry *,int);
38 int (*rmdir) (struct inode *,struct dentry *);
39 int (*mknod) (struct inode *,struct dentry *,int,dev_t);
40 int (*rename) (struct inode *, struct dentry *,
41 struct inode *, struct dentry *);
42 int (*readlink) (struct dentry *, char __user *,int);
43 int (*follow_link) (struct dentry *, struct nameidata *);
44 void (*truncate) (struct inode *);
45 int (*permission) (struct inode *, int, struct nameidata *);
46 int (*setattr) (struct dentry *, struct iattr *);
47 int (*getattr) (struct vfsmount *, struct dentry *, struct kstat *);
48 int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
49 ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t);
50 ssize_t (*listxattr) (struct dentry *, char *, size_t);
51 int (*removexattr) (struct dentry *, const char *);
54 all may block, none have BKL
63 rmdir: yes (both) (see below)
64 rename: yes (all) (see below)
67 truncate: yes (see below)
75 Additionally, ->rmdir(), ->unlink() and ->rename() have ->i_sem on
77 cross-directory ->rename() has (per-superblock) ->s_vfs_rename_sem.
78 ->truncate() is never called directly - it's a callback, not a
79 method. It's called by vmtruncate() - library function normally used by
80 ->setattr(). Locking information above applies to that call (i.e. is
81 inherited from ->setattr() - vmtruncate() is used when ATTR_SIZE had been
84 See Documentation/filesystems/directory-locking for more detailed discussion
85 of the locking scheme for directory operations.
87 --------------------------- super_operations ---------------------------
89 struct inode *(*alloc_inode)(struct super_block *sb);
90 void (*destroy_inode)(struct inode *);
91 void (*read_inode) (struct inode *);
92 void (*dirty_inode) (struct inode *);
93 int (*write_inode) (struct inode *, int);
94 void (*put_inode) (struct inode *);
95 void (*drop_inode) (struct inode *);
96 void (*delete_inode) (struct inode *);
97 void (*put_super) (struct super_block *);
98 void (*write_super) (struct super_block *);
99 int (*sync_fs)(struct super_block *sb, int wait);
100 void (*write_super_lockfs) (struct super_block *);
101 void (*unlockfs) (struct super_block *);
102 int (*statfs) (struct super_block *, struct kstatfs *);
103 int (*remount_fs) (struct super_block *, int *, char *);
104 void (*clear_inode) (struct inode *);
105 void (*umount_begin) (struct super_block *);
106 int (*show_options)(struct seq_file *, struct vfsmount *);
107 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
108 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
113 alloc_inode: no no no
115 read_inode: no (see below)
116 dirty_inode: no (must not sleep)
119 drop_inode: no !!!inode_lock!!!
121 put_super: yes yes no
122 write_super: no yes read
124 write_super_lockfs: ?
127 remount_fs: no yes maybe (see below)
129 umount_begin: yes no no
130 show_options: no (vfsmount->sem)
131 quota_read: no no no (see below)
132 quota_write: no no no (see below)
134 ->read_inode() is not a method - it's a callback used in iget().
135 ->remount_fs() will have the s_umount lock if it's already mounted.
136 When called from get_sb_single, it does NOT have the s_umount lock.
137 ->quota_read() and ->quota_write() functions are both guaranteed to
138 be the only ones operating on the quota file by the quota code (via
139 dqio_sem) (unless an admin really wants to screw up something and
140 writes to quota files with quotas on). For other details about locking
141 see also dquot_operations section.
143 --------------------------- file_system_type ---------------------------
145 struct super_block *(*get_sb) (struct file_system_type *, int,
146 const char *, void *);
147 void (*kill_sb) (struct super_block *);
153 ->get_sb() returns error or a locked superblock (exclusive on ->s_umount).
154 ->kill_sb() takes a write-locked superblock, does all shutdown work on it,
155 unlocks and drops the reference.
157 --------------------------- address_space_operations --------------------------
159 int (*writepage)(struct page *page, struct writeback_control *wbc);
160 int (*readpage)(struct file *, struct page *);
161 int (*sync_page)(struct page *);
162 int (*writepages)(struct address_space *, struct writeback_control *);
163 int (*set_page_dirty)(struct page *page);
164 int (*readpages)(struct file *filp, struct address_space *mapping,
165 struct list_head *pages, unsigned nr_pages);
166 int (*prepare_write)(struct file *, struct page *, unsigned, unsigned);
167 int (*commit_write)(struct file *, struct page *, unsigned, unsigned);
168 sector_t (*bmap)(struct address_space *, sector_t);
169 int (*invalidatepage) (struct page *, unsigned long);
170 int (*releasepage) (struct page *, int);
171 int (*direct_IO)(int, struct kiocb *, const struct iovec *iov,
172 loff_t offset, unsigned long nr_segs);
175 All except set_page_dirty may block
178 writepage: no yes, unlocks (see below)
179 readpage: no yes, unlocks
184 prepare_write: no yes
187 invalidatepage: no yes
191 ->prepare_write(), ->commit_write(), ->sync_page() and ->readpage()
192 may be called from the request handler (/dev/loop).
194 ->readpage() unlocks the page, either synchronously or via I/O
197 ->readpages() populates the pagecache with the passed pages and starts
198 I/O against them. They come unlocked upon I/O completion.
200 ->writepage() is used for two purposes: for "memory cleansing" and for
201 "sync". These are quite different operations and the behaviour may differ
202 depending upon the mode.
204 If writepage is called for sync (wbc->sync_mode != WBC_SYNC_NONE) then
205 it *must* start I/O against the page, even if that would involve
206 blocking on in-progress I/O.
208 If writepage is called for memory cleansing (sync_mode ==
209 WBC_SYNC_NONE) then its role is to get as much writeout underway as
210 possible. So writepage should try to avoid blocking against
211 currently-in-progress I/O.
213 If the filesystem is not called for "sync" and it determines that it
214 would need to block against in-progress I/O to be able to start new I/O
215 against the page the filesystem should redirty the page with
216 redirty_page_for_writepage(), then unlock the page and return zero.
217 This may also be done to avoid internal deadlocks, but rarely.
219 If the filesytem is called for sync then it must wait on any
220 in-progress I/O and then start new I/O.
222 The filesystem should unlock the page synchronously, before returning to the
223 caller, unless ->writepage() returns special WRITEPAGE_ACTIVATE
224 value. WRITEPAGE_ACTIVATE means that page cannot really be written out
225 currently, and VM should stop calling ->writepage() on this page for some
226 time. VM does this by moving page to the head of the active list, hence the
229 Unless the filesystem is going to redirty_page_for_writepage(), unlock the page
230 and return zero, writepage *must* run set_page_writeback() against the page,
231 followed by unlocking it. Once set_page_writeback() has been run against the
232 page, write I/O can be submitted and the write I/O completion handler must run
233 end_page_writeback() once the I/O is complete. If no I/O is submitted, the
234 filesystem must run end_page_writeback() against the page before returning from
237 That is: after 2.5.12, pages which are under writeout are *not* locked. Note,
238 if the filesystem needs the page to be locked during writeout, that is ok, too,
239 the page is allowed to be unlocked at any point in time between the calls to
240 set_page_writeback() and end_page_writeback().
242 Note, failure to run either redirty_page_for_writepage() or the combination of
243 set_page_writeback()/end_page_writeback() on a page submitted to writepage
244 will leave the page itself marked clean but it will be tagged as dirty in the
245 radix tree. This incoherency can lead to all sorts of hard-to-debug problems
246 in the filesystem like having dirty inodes at umount and losing written data.
248 ->sync_page() locking rules are not well-defined - usually it is called
249 with lock on page, but that is not guaranteed. Considering the currently
250 existing instances of this method ->sync_page() itself doesn't look
253 ->writepages() is used for periodic writeback and for syscall-initiated
254 sync operations. The address_space should start I/O against at least
255 *nr_to_write pages. *nr_to_write must be decremented for each page which is
256 written. The address_space implementation may write more (or less) pages
257 than *nr_to_write asks for, but it should try to be reasonably close. If
258 nr_to_write is NULL, all dirty pages must be written.
260 writepages should _only_ write pages which are present on
263 ->set_page_dirty() is called from various places in the kernel
264 when the target page is marked as needing writeback. It may be called
265 under spinlock (it cannot block) and is sometimes called with the page
268 ->bmap() is currently used by legacy ioctl() (FIBMAP) provided by some
269 filesystems and by the swapper. The latter will eventually go away. All
270 instances do not actually need the BKL. Please, keep it that way and don't
273 ->invalidatepage() is called when the filesystem must attempt to drop
274 some or all of the buffers from the page when it is being truncated. It
275 returns zero on success. If ->invalidatepage is zero, the kernel uses
276 block_invalidatepage() instead.
278 ->releasepage() is called when the kernel is about to try to drop the
279 buffers from the page in preparation for freeing it. It returns zero to
280 indicate that the buffers are (or may be) freeable. If ->releasepage is zero,
281 the kernel assumes that the fs has no private interest in the buffers.
283 Note: currently almost all instances of address_space methods are
284 using BKL for internal serialization and that's one of the worst sources
285 of contention. Normally they are calling library functions (in fs/buffer.c)
286 and pass foo_get_block() as a callback (on local block-based filesystems,
287 indeed). BKL is not needed for library stuff and is usually taken by
288 foo_get_block(). It's an overkill, since block bitmaps can be protected by
289 internal fs locking and real critical areas are much smaller than the areas
290 filesystems protect now.
292 ----------------------- file_lock_operations ------------------------------
294 void (*fl_insert)(struct file_lock *); /* lock insertion callback */
295 void (*fl_remove)(struct file_lock *); /* lock removal callback */
296 void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
297 void (*fl_release_private)(struct file_lock *);
305 fl_release_private: yes yes
307 ----------------------- lock_manager_operations ---------------------------
309 int (*fl_compare_owner)(struct file_lock *, struct file_lock *);
310 void (*fl_notify)(struct file_lock *); /* unblock callback */
311 void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
312 void (*fl_release_private)(struct file_lock *);
313 void (*fl_break)(struct file_lock *); /* break_lease callback */
317 fl_compare_owner: yes no
320 fl_release_private: yes yes
323 Currently only NFSD and NLM provide instances of this class. None of the
324 them block. If you have out-of-tree instances - please, show up. Locking
325 in that area will change.
326 --------------------------- buffer_head -----------------------------------
328 void (*b_end_io)(struct buffer_head *bh, int uptodate);
331 called from interrupts. In other words, extreme care is needed here.
332 bh is locked, but that's all warranties we have here. Currently only RAID1,
333 highmem, fs/buffer.c, and fs/ntfs/aops.c are providing these. Block devices
334 call this method upon the IO completion.
336 --------------------------- block_device_operations -----------------------
338 int (*open) (struct inode *, struct file *);
339 int (*release) (struct inode *, struct file *);
340 int (*ioctl) (struct inode *, struct file *, unsigned, unsigned long);
341 int (*media_changed) (struct gendisk *);
342 int (*revalidate_disk) (struct gendisk *);
350 revalidate_disk: no no
352 The last two are called only from check_disk_change().
354 --------------------------- file_operations -------------------------------
356 loff_t (*llseek) (struct file *, loff_t, int);
357 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
358 ssize_t (*aio_read) (struct kiocb *, char __user *, size_t, loff_t);
359 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
360 ssize_t (*aio_write) (struct kiocb *, const char __user *, size_t,
362 int (*readdir) (struct file *, void *, filldir_t);
363 unsigned int (*poll) (struct file *, struct poll_table_struct *);
364 int (*ioctl) (struct inode *, struct file *, unsigned int,
366 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
367 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
368 int (*mmap) (struct file *, struct vm_area_struct *);
369 int (*open) (struct inode *, struct file *);
370 int (*flush) (struct file *);
371 int (*release) (struct inode *, struct file *);
372 int (*fsync) (struct file *, struct dentry *, int datasync);
373 int (*aio_fsync) (struct kiocb *, int datasync);
374 int (*fasync) (int, struct file *, int);
375 int (*lock) (struct file *, int, struct file_lock *);
376 ssize_t (*readv) (struct file *, const struct iovec *, unsigned long,
378 ssize_t (*writev) (struct file *, const struct iovec *, unsigned long,
380 ssize_t (*sendfile) (struct file *, loff_t *, size_t, read_actor_t,
382 ssize_t (*sendpage) (struct file *, struct page *, int, size_t,
384 unsigned long (*get_unmapped_area)(struct file *, unsigned long,
385 unsigned long, unsigned long, unsigned long);
386 int (*check_flags)(int);
387 int (*dir_notify)(struct file *, unsigned long);
391 All except ->poll() may block.
393 llseek: no (see below)
400 ioctl: yes (see below)
401 unlocked_ioctl: no (see below)
404 open: maybe (see below)
407 fsync: no (see below)
409 fasync: yes (see below)
415 get_unmapped_area: no
419 ->llseek() locking has moved from llseek to the individual llseek
420 implementations. If your fs is not using generic_file_llseek, you
421 need to acquire and release the appropriate locks in your ->llseek().
422 For many filesystems, it is probably safe to acquire the inode
423 semaphore. Note some filesystems (i.e. remote ones) provide no
424 protection for i_size so you will need to use the BKL.
426 ->open() locking is in-transit: big lock partially moved into the methods.
427 The only exception is ->open() in the instances of file_operations that never
428 end up in ->i_fop/->proc_fops, i.e. ones that belong to character devices
429 (chrdev_open() takes lock before replacing ->f_op and calling the secondary
430 method. As soon as we fix the handling of module reference counters all
431 instances of ->open() will be called without the BKL.
433 Note: ext2_release() was *the* source of contention on fs-intensive
434 loads and dropping BKL on ->release() helps to get rid of that (we still
435 grab BKL for cases when we close a file that had been opened r/w, but that
436 can and should be done using the internal locking with smaller critical areas).
437 Current worst offender is ext2_get_block()...
439 ->fasync() is a mess. This area needs a big cleanup and that will probably
442 ->readdir() and ->ioctl() on directories must be changed. Ideally we would
443 move ->readdir() to inode_operations and use a separate method for directory
444 ->ioctl() or kill the latter completely. One of the problems is that for
445 anything that resembles union-mount we won't have a struct file for all
446 components. And there are other reasons why the current interface is a mess...
448 ->ioctl() on regular files is superceded by the ->unlocked_ioctl() that
449 doesn't take the BKL.
451 ->read on directories probably must go away - we should just enforce -EISDIR
452 in sys_read() and friends.
454 ->fsync() has i_sem on inode.
456 --------------------------- dquot_operations -------------------------------
458 int (*initialize) (struct inode *, int);
459 int (*drop) (struct inode *);
460 int (*alloc_space) (struct inode *, qsize_t, int);
461 int (*alloc_inode) (const struct inode *, unsigned long);
462 int (*free_space) (struct inode *, qsize_t);
463 int (*free_inode) (const struct inode *, unsigned long);
464 int (*transfer) (struct inode *, struct iattr *);
465 int (*write_dquot) (struct dquot *);
466 int (*acquire_dquot) (struct dquot *);
467 int (*release_dquot) (struct dquot *);
468 int (*mark_dirty) (struct dquot *);
469 int (*write_info) (struct super_block *, int);
471 These operations are intended to be more or less wrapping functions that ensure
472 a proper locking wrt the filesystem and call the generic quota operations.
474 What filesystem should expect from the generic quota functions:
476 FS recursion Held locks when called
477 initialize: yes maybe dqonoff_sem
479 alloc_space: ->mark_dirty() -
480 alloc_inode: ->mark_dirty() -
481 free_space: ->mark_dirty() -
482 free_inode: ->mark_dirty() -
484 write_dquot: yes dqonoff_sem or dqptr_sem
485 acquire_dquot: yes dqonoff_sem or dqptr_sem
486 release_dquot: yes dqonoff_sem or dqptr_sem
488 write_info: yes dqonoff_sem
490 FS recursion means calling ->quota_read() and ->quota_write() from superblock
493 ->alloc_space(), ->alloc_inode(), ->free_space(), ->free_inode() are called
494 only directly by the filesystem and do not call any fs functions only
495 the ->mark_dirty() operation.
497 More details about quota locking can be found in fs/dquot.c.
499 --------------------------- vm_operations_struct -----------------------------
501 void (*open)(struct vm_area_struct*);
502 void (*close)(struct vm_area_struct*);
503 struct page *(*nopage)(struct vm_area_struct*, unsigned long, int *);
511 ================================================================================
514 (if you break something or notice that it is broken and do not fix it yourself
515 - at least put it here)
517 ipc/shm.c::shm_delete() - may need BKL.
518 ->read() and ->write() in many drivers are (probably) missing BKL.
519 drivers/sgi/char/graphics.c::sgi_graphics_nopage() - may need BKL.