blob: d866551be03790c20a00b5f96c41f9226260bcfa [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001 The text below describes the locking rules for VFS-related methods.
2It is (believed to be) up-to-date. *Please*, if you change anything in
3prototypes or locking protocols - update this file. And update the relevant
4instances in the tree, don't leave that to maintainers of filesystems/devices/
5etc. At the very least, put the list of dubious cases in the end of this file.
6Don't turn it into log - maintainers of out-of-the-tree code are supposed to
7be able to use diff(1).
8 Thing currently missing here: socket operations. Alexey?
9
10--------------------------- dentry_operations --------------------------
11prototypes:
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 *);
Eric Dumazetc23fbb62007-05-08 00:26:18 -070018 char *(*d_dname)((struct dentry *dentry, char *buffer, int buflen);
Linus Torvalds1da177e2005-04-16 15:20:36 -070019
20locking rules:
21 none have BKL
22 dcache_lock rename_lock ->d_lock may block
23d_revalidate: no no no yes
24d_hash no no no yes
25d_compare: no yes no no
26d_delete: yes no yes no
27d_release: no no no yes
28d_iput: no no no yes
Eric Dumazetc23fbb62007-05-08 00:26:18 -070029d_dname: no no no no
Linus Torvalds1da177e2005-04-16 15:20:36 -070030
31--------------------------- inode_operations ---------------------------
32prototypes:
33 int (*create) (struct inode *,struct dentry *,int, struct nameidata *);
34 struct dentry * (*lookup) (struct inode *,struct dentry *, struct nameid
35ata *);
36 int (*link) (struct dentry *,struct inode *,struct dentry *);
37 int (*unlink) (struct inode *,struct dentry *);
38 int (*symlink) (struct inode *,struct dentry *,const char *);
39 int (*mkdir) (struct inode *,struct dentry *,int);
40 int (*rmdir) (struct inode *,struct dentry *);
41 int (*mknod) (struct inode *,struct dentry *,int,dev_t);
42 int (*rename) (struct inode *, struct dentry *,
43 struct inode *, struct dentry *);
44 int (*readlink) (struct dentry *, char __user *,int);
45 int (*follow_link) (struct dentry *, struct nameidata *);
46 void (*truncate) (struct inode *);
47 int (*permission) (struct inode *, int, struct nameidata *);
48 int (*setattr) (struct dentry *, struct iattr *);
49 int (*getattr) (struct vfsmount *, struct dentry *, struct kstat *);
50 int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
51 ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t);
52 ssize_t (*listxattr) (struct dentry *, char *, size_t);
53 int (*removexattr) (struct dentry *, const char *);
54
55locking rules:
56 all may block, none have BKL
Artem Bityutskiya7bc02f2007-05-09 07:53:16 +020057 i_mutex(inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -070058lookup: yes
59create: yes
60link: yes (both)
61mknod: yes
62symlink: yes
63mkdir: yes
64unlink: yes (both)
65rmdir: yes (both) (see below)
66rename: yes (all) (see below)
67readlink: no
68follow_link: no
69truncate: yes (see below)
70setattr: yes
71permission: no
72getattr: no
73setxattr: yes
74getxattr: no
75listxattr: no
76removexattr: yes
Artem Bityutskiya7bc02f2007-05-09 07:53:16 +020077 Additionally, ->rmdir(), ->unlink() and ->rename() have ->i_mutex on
Linus Torvalds1da177e2005-04-16 15:20:36 -070078victim.
79 cross-directory ->rename() has (per-superblock) ->s_vfs_rename_sem.
80 ->truncate() is never called directly - it's a callback, not a
81method. It's called by vmtruncate() - library function normally used by
82->setattr(). Locking information above applies to that call (i.e. is
83inherited from ->setattr() - vmtruncate() is used when ATTR_SIZE had been
84passed).
85
86See Documentation/filesystems/directory-locking for more detailed discussion
87of the locking scheme for directory operations.
88
89--------------------------- super_operations ---------------------------
90prototypes:
91 struct inode *(*alloc_inode)(struct super_block *sb);
92 void (*destroy_inode)(struct inode *);
93 void (*read_inode) (struct inode *);
94 void (*dirty_inode) (struct inode *);
95 int (*write_inode) (struct inode *, int);
96 void (*put_inode) (struct inode *);
97 void (*drop_inode) (struct inode *);
98 void (*delete_inode) (struct inode *);
99 void (*put_super) (struct super_block *);
100 void (*write_super) (struct super_block *);
101 int (*sync_fs)(struct super_block *sb, int wait);
102 void (*write_super_lockfs) (struct super_block *);
103 void (*unlockfs) (struct super_block *);
David Howells726c3342006-06-23 02:02:58 -0700104 int (*statfs) (struct dentry *, struct kstatfs *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 int (*remount_fs) (struct super_block *, int *, char *);
106 void (*clear_inode) (struct inode *);
107 void (*umount_begin) (struct super_block *);
108 int (*show_options)(struct seq_file *, struct vfsmount *);
109 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
110 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
111
112locking rules:
113 All may block.
114 BKL s_lock s_umount
115alloc_inode: no no no
116destroy_inode: no
117read_inode: no (see below)
118dirty_inode: no (must not sleep)
119write_inode: no
120put_inode: no
121drop_inode: no !!!inode_lock!!!
122delete_inode: no
123put_super: yes yes no
124write_super: no yes read
125sync_fs: no no read
126write_super_lockfs: ?
127unlockfs: ?
128statfs: no no no
Vasily Averin70888bd2006-12-06 20:37:23 -0800129remount_fs: yes yes maybe (see below)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130clear_inode: no
131umount_begin: yes no no
132show_options: no (vfsmount->sem)
133quota_read: no no no (see below)
134quota_write: no no no (see below)
135
136->read_inode() is not a method - it's a callback used in iget().
137->remount_fs() will have the s_umount lock if it's already mounted.
138When called from get_sb_single, it does NOT have the s_umount lock.
139->quota_read() and ->quota_write() functions are both guaranteed to
140be the only ones operating on the quota file by the quota code (via
141dqio_sem) (unless an admin really wants to screw up something and
142writes to quota files with quotas on). For other details about locking
143see also dquot_operations section.
144
145--------------------------- file_system_type ---------------------------
146prototypes:
Jonathan Corbet5d8b2eb2006-07-10 04:44:07 -0700147 int (*get_sb) (struct file_system_type *, int,
148 const char *, void *, struct vfsmount *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 void (*kill_sb) (struct super_block *);
150locking rules:
151 may block BKL
152get_sb yes yes
153kill_sb yes yes
154
David Howells454e2392006-06-23 02:02:57 -0700155->get_sb() returns error or 0 with locked superblock attached to the vfsmount
156(exclusive on ->s_umount).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157->kill_sb() takes a write-locked superblock, does all shutdown work on it,
158unlocks and drops the reference.
159
160--------------------------- address_space_operations --------------------------
161prototypes:
162 int (*writepage)(struct page *page, struct writeback_control *wbc);
163 int (*readpage)(struct file *, struct page *);
164 int (*sync_page)(struct page *);
165 int (*writepages)(struct address_space *, struct writeback_control *);
166 int (*set_page_dirty)(struct page *page);
167 int (*readpages)(struct file *filp, struct address_space *mapping,
168 struct list_head *pages, unsigned nr_pages);
169 int (*prepare_write)(struct file *, struct page *, unsigned, unsigned);
170 int (*commit_write)(struct file *, struct page *, unsigned, unsigned);
171 sector_t (*bmap)(struct address_space *, sector_t);
172 int (*invalidatepage) (struct page *, unsigned long);
173 int (*releasepage) (struct page *, int);
174 int (*direct_IO)(int, struct kiocb *, const struct iovec *iov,
175 loff_t offset, unsigned long nr_segs);
Trond Myklebuste3db7692007-01-10 23:15:39 -0800176 int (*launder_page) (struct page *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177
178locking rules:
179 All except set_page_dirty may block
180
181 BKL PageLocked(page)
182writepage: no yes, unlocks (see below)
183readpage: no yes, unlocks
184sync_page: no maybe
185writepages: no
186set_page_dirty no no
187readpages: no
188prepare_write: no yes
189commit_write: no yes
190bmap: yes
191invalidatepage: no yes
192releasepage: no yes
193direct_IO: no
Trond Myklebuste3db7692007-01-10 23:15:39 -0800194launder_page: no yes
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195
196 ->prepare_write(), ->commit_write(), ->sync_page() and ->readpage()
197may be called from the request handler (/dev/loop).
198
199 ->readpage() unlocks the page, either synchronously or via I/O
200completion.
201
202 ->readpages() populates the pagecache with the passed pages and starts
203I/O against them. They come unlocked upon I/O completion.
204
205 ->writepage() is used for two purposes: for "memory cleansing" and for
206"sync". These are quite different operations and the behaviour may differ
207depending upon the mode.
208
209If writepage is called for sync (wbc->sync_mode != WBC_SYNC_NONE) then
210it *must* start I/O against the page, even if that would involve
211blocking on in-progress I/O.
212
213If writepage is called for memory cleansing (sync_mode ==
214WBC_SYNC_NONE) then its role is to get as much writeout underway as
215possible. So writepage should try to avoid blocking against
216currently-in-progress I/O.
217
218If the filesystem is not called for "sync" and it determines that it
219would need to block against in-progress I/O to be able to start new I/O
220against the page the filesystem should redirty the page with
221redirty_page_for_writepage(), then unlock the page and return zero.
222This may also be done to avoid internal deadlocks, but rarely.
223
224If the filesytem is called for sync then it must wait on any
225in-progress I/O and then start new I/O.
226
Nikita Danilov20546062005-05-01 08:58:37 -0700227The filesystem should unlock the page synchronously, before returning to the
228caller, unless ->writepage() returns special WRITEPAGE_ACTIVATE
229value. WRITEPAGE_ACTIVATE means that page cannot really be written out
230currently, and VM should stop calling ->writepage() on this page for some
231time. VM does this by moving page to the head of the active list, hence the
232name.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233
234Unless the filesystem is going to redirty_page_for_writepage(), unlock the page
235and return zero, writepage *must* run set_page_writeback() against the page,
236followed by unlocking it. Once set_page_writeback() has been run against the
237page, write I/O can be submitted and the write I/O completion handler must run
238end_page_writeback() once the I/O is complete. If no I/O is submitted, the
239filesystem must run end_page_writeback() against the page before returning from
240writepage.
241
242That is: after 2.5.12, pages which are under writeout are *not* locked. Note,
243if the filesystem needs the page to be locked during writeout, that is ok, too,
244the page is allowed to be unlocked at any point in time between the calls to
245set_page_writeback() and end_page_writeback().
246
247Note, failure to run either redirty_page_for_writepage() or the combination of
248set_page_writeback()/end_page_writeback() on a page submitted to writepage
249will leave the page itself marked clean but it will be tagged as dirty in the
250radix tree. This incoherency can lead to all sorts of hard-to-debug problems
251in the filesystem like having dirty inodes at umount and losing written data.
252
253 ->sync_page() locking rules are not well-defined - usually it is called
254with lock on page, but that is not guaranteed. Considering the currently
255existing instances of this method ->sync_page() itself doesn't look
256well-defined...
257
258 ->writepages() is used for periodic writeback and for syscall-initiated
259sync operations. The address_space should start I/O against at least
260*nr_to_write pages. *nr_to_write must be decremented for each page which is
261written. The address_space implementation may write more (or less) pages
262than *nr_to_write asks for, but it should try to be reasonably close. If
263nr_to_write is NULL, all dirty pages must be written.
264
265writepages should _only_ write pages which are present on
266mapping->io_pages.
267
268 ->set_page_dirty() is called from various places in the kernel
269when the target page is marked as needing writeback. It may be called
270under spinlock (it cannot block) and is sometimes called with the page
271not locked.
272
273 ->bmap() is currently used by legacy ioctl() (FIBMAP) provided by some
274filesystems and by the swapper. The latter will eventually go away. All
275instances do not actually need the BKL. Please, keep it that way and don't
276breed new callers.
277
278 ->invalidatepage() is called when the filesystem must attempt to drop
279some or all of the buffers from the page when it is being truncated. It
280returns zero on success. If ->invalidatepage is zero, the kernel uses
281block_invalidatepage() instead.
282
283 ->releasepage() is called when the kernel is about to try to drop the
284buffers from the page in preparation for freeing it. It returns zero to
285indicate that the buffers are (or may be) freeable. If ->releasepage is zero,
286the kernel assumes that the fs has no private interest in the buffers.
287
Trond Myklebuste3db7692007-01-10 23:15:39 -0800288 ->launder_page() may be called prior to releasing a page if
289it is still found to be dirty. It returns zero if the page was successfully
290cleaned, or an error value if not. Note that in order to prevent the page
291getting mapped back in and redirtied, it needs to be kept locked
292across the entire operation.
293
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 Note: currently almost all instances of address_space methods are
295using BKL for internal serialization and that's one of the worst sources
296of contention. Normally they are calling library functions (in fs/buffer.c)
297and pass foo_get_block() as a callback (on local block-based filesystems,
298indeed). BKL is not needed for library stuff and is usually taken by
299foo_get_block(). It's an overkill, since block bitmaps can be protected by
300internal fs locking and real critical areas are much smaller than the areas
301filesystems protect now.
302
303----------------------- file_lock_operations ------------------------------
304prototypes:
305 void (*fl_insert)(struct file_lock *); /* lock insertion callback */
306 void (*fl_remove)(struct file_lock *); /* lock removal callback */
307 void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
308 void (*fl_release_private)(struct file_lock *);
309
310
311locking rules:
312 BKL may block
313fl_insert: yes no
314fl_remove: yes no
315fl_copy_lock: yes no
316fl_release_private: yes yes
317
318----------------------- lock_manager_operations ---------------------------
319prototypes:
320 int (*fl_compare_owner)(struct file_lock *, struct file_lock *);
321 void (*fl_notify)(struct file_lock *); /* unblock callback */
322 void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
323 void (*fl_release_private)(struct file_lock *);
324 void (*fl_break)(struct file_lock *); /* break_lease callback */
325
326locking rules:
327 BKL may block
328fl_compare_owner: yes no
329fl_notify: yes no
330fl_copy_lock: yes no
331fl_release_private: yes yes
332fl_break: yes no
333
334 Currently only NFSD and NLM provide instances of this class. None of the
335them block. If you have out-of-tree instances - please, show up. Locking
336in that area will change.
337--------------------------- buffer_head -----------------------------------
338prototypes:
339 void (*b_end_io)(struct buffer_head *bh, int uptodate);
340
341locking rules:
342 called from interrupts. In other words, extreme care is needed here.
343bh is locked, but that's all warranties we have here. Currently only RAID1,
344highmem, fs/buffer.c, and fs/ntfs/aops.c are providing these. Block devices
345call this method upon the IO completion.
346
347--------------------------- block_device_operations -----------------------
348prototypes:
349 int (*open) (struct inode *, struct file *);
350 int (*release) (struct inode *, struct file *);
351 int (*ioctl) (struct inode *, struct file *, unsigned, unsigned long);
352 int (*media_changed) (struct gendisk *);
353 int (*revalidate_disk) (struct gendisk *);
354
355locking rules:
356 BKL bd_sem
357open: yes yes
358release: yes yes
359ioctl: yes no
360media_changed: no no
361revalidate_disk: no no
362
363The last two are called only from check_disk_change().
364
365--------------------------- file_operations -------------------------------
366prototypes:
367 loff_t (*llseek) (struct file *, loff_t, int);
368 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700370 ssize_t (*aio_read) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
371 ssize_t (*aio_write) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 int (*readdir) (struct file *, void *, filldir_t);
373 unsigned int (*poll) (struct file *, struct poll_table_struct *);
374 int (*ioctl) (struct inode *, struct file *, unsigned int,
375 unsigned long);
376 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
377 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
378 int (*mmap) (struct file *, struct vm_area_struct *);
379 int (*open) (struct inode *, struct file *);
380 int (*flush) (struct file *);
381 int (*release) (struct inode *, struct file *);
382 int (*fsync) (struct file *, struct dentry *, int datasync);
383 int (*aio_fsync) (struct kiocb *, int datasync);
384 int (*fasync) (int, struct file *, int);
385 int (*lock) (struct file *, int, struct file_lock *);
386 ssize_t (*readv) (struct file *, const struct iovec *, unsigned long,
387 loff_t *);
388 ssize_t (*writev) (struct file *, const struct iovec *, unsigned long,
389 loff_t *);
390 ssize_t (*sendfile) (struct file *, loff_t *, size_t, read_actor_t,
391 void __user *);
392 ssize_t (*sendpage) (struct file *, struct page *, int, size_t,
393 loff_t *, int);
394 unsigned long (*get_unmapped_area)(struct file *, unsigned long,
395 unsigned long, unsigned long, unsigned long);
396 int (*check_flags)(int);
397 int (*dir_notify)(struct file *, unsigned long);
398};
399
400locking rules:
401 All except ->poll() may block.
402 BKL
403llseek: no (see below)
404read: no
405aio_read: no
406write: no
407aio_write: no
408readdir: no
409poll: no
410ioctl: yes (see below)
411unlocked_ioctl: no (see below)
412compat_ioctl: no
413mmap: no
414open: maybe (see below)
415flush: no
416release: no
417fsync: no (see below)
418aio_fsync: no
419fasync: yes (see below)
420lock: yes
421readv: no
422writev: no
423sendfile: no
424sendpage: no
425get_unmapped_area: no
426check_flags: no
427dir_notify: no
428
429->llseek() locking has moved from llseek to the individual llseek
430implementations. If your fs is not using generic_file_llseek, you
431need to acquire and release the appropriate locks in your ->llseek().
432For many filesystems, it is probably safe to acquire the inode
433semaphore. Note some filesystems (i.e. remote ones) provide no
434protection for i_size so you will need to use the BKL.
435
436->open() locking is in-transit: big lock partially moved into the methods.
437The only exception is ->open() in the instances of file_operations that never
438end up in ->i_fop/->proc_fops, i.e. ones that belong to character devices
439(chrdev_open() takes lock before replacing ->f_op and calling the secondary
440method. As soon as we fix the handling of module reference counters all
441instances of ->open() will be called without the BKL.
442
443Note: ext2_release() was *the* source of contention on fs-intensive
444loads and dropping BKL on ->release() helps to get rid of that (we still
445grab BKL for cases when we close a file that had been opened r/w, but that
446can and should be done using the internal locking with smaller critical areas).
447Current worst offender is ext2_get_block()...
448
449->fasync() is a mess. This area needs a big cleanup and that will probably
450affect locking.
451
452->readdir() and ->ioctl() on directories must be changed. Ideally we would
453move ->readdir() to inode_operations and use a separate method for directory
454->ioctl() or kill the latter completely. One of the problems is that for
455anything that resembles union-mount we won't have a struct file for all
456components. And there are other reasons why the current interface is a mess...
457
458->ioctl() on regular files is superceded by the ->unlocked_ioctl() that
459doesn't take the BKL.
460
461->read on directories probably must go away - we should just enforce -EISDIR
462in sys_read() and friends.
463
Artem Bityutskiya7bc02f2007-05-09 07:53:16 +0200464->fsync() has i_mutex on inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
466--------------------------- dquot_operations -------------------------------
467prototypes:
468 int (*initialize) (struct inode *, int);
469 int (*drop) (struct inode *);
470 int (*alloc_space) (struct inode *, qsize_t, int);
471 int (*alloc_inode) (const struct inode *, unsigned long);
472 int (*free_space) (struct inode *, qsize_t);
473 int (*free_inode) (const struct inode *, unsigned long);
474 int (*transfer) (struct inode *, struct iattr *);
475 int (*write_dquot) (struct dquot *);
476 int (*acquire_dquot) (struct dquot *);
477 int (*release_dquot) (struct dquot *);
478 int (*mark_dirty) (struct dquot *);
479 int (*write_info) (struct super_block *, int);
480
481These operations are intended to be more or less wrapping functions that ensure
482a proper locking wrt the filesystem and call the generic quota operations.
483
484What filesystem should expect from the generic quota functions:
485
486 FS recursion Held locks when called
487initialize: yes maybe dqonoff_sem
488drop: yes -
489alloc_space: ->mark_dirty() -
490alloc_inode: ->mark_dirty() -
491free_space: ->mark_dirty() -
492free_inode: ->mark_dirty() -
493transfer: yes -
494write_dquot: yes dqonoff_sem or dqptr_sem
495acquire_dquot: yes dqonoff_sem or dqptr_sem
496release_dquot: yes dqonoff_sem or dqptr_sem
497mark_dirty: no -
498write_info: yes dqonoff_sem
499
500FS recursion means calling ->quota_read() and ->quota_write() from superblock
501operations.
502
503->alloc_space(), ->alloc_inode(), ->free_space(), ->free_inode() are called
504only directly by the filesystem and do not call any fs functions only
505the ->mark_dirty() operation.
506
507More details about quota locking can be found in fs/dquot.c.
508
509--------------------------- vm_operations_struct -----------------------------
510prototypes:
511 void (*open)(struct vm_area_struct*);
512 void (*close)(struct vm_area_struct*);
513 struct page *(*nopage)(struct vm_area_struct*, unsigned long, int *);
514
515locking rules:
516 BKL mmap_sem
517open: no yes
518close: no yes
519nopage: no yes
520
521================================================================================
522 Dubious stuff
523
524(if you break something or notice that it is broken and do not fix it yourself
525- at least put it here)
526
527ipc/shm.c::shm_delete() - may need BKL.
528->read() and ->write() in many drivers are (probably) missing BKL.
529drivers/sgi/char/graphics.c::sgi_graphics_nopage() - may need BKL.