5 The text below describes the locking rules for VFS-related methods.
6 It is (believed to be) up-to-date. *Please*, if you change anything in
7 prototypes or locking protocols - update this file. And update the relevant
8 instances in the tree, don't leave that to maintainers of filesystems/devices/
9 etc. At the very least, put the list of dubious cases in the end of this file.
10 Don't turn it into log - maintainers of out-of-the-tree code are supposed to
11 be able to use diff(1).
13 Thing currently missing here: socket operations. Alexey?
20 int (*d_revalidate)(struct dentry *, unsigned int);
21 int (*d_weak_revalidate)(struct dentry *, unsigned int);
22 int (*d_hash)(const struct dentry *, struct qstr *);
23 int (*d_compare)(const struct dentry *,
24 unsigned int, const char *, const struct qstr *);
25 int (*d_delete)(struct dentry *);
26 int (*d_init)(struct dentry *);
27 void (*d_release)(struct dentry *);
28 void (*d_iput)(struct dentry *, struct inode *);
29 char *(*d_dname)((struct dentry *dentry, char *buffer, int buflen);
30 struct vfsmount *(*d_automount)(struct path *path);
31 int (*d_manage)(const struct path *, bool);
32 struct dentry *(*d_real)(struct dentry *, const struct inode *);
36 ================== =========== ======== ============== ========
37 ops rename_lock ->d_lock may block rcu-walk
38 ================== =========== ======== ============== ========
39 d_revalidate: no no yes (ref-walk) maybe
40 d_weak_revalidate: no no yes no
42 d_compare: yes no no maybe
43 d_delete: no yes no no
45 d_release: no no yes no
49 d_automount: no no yes no
50 d_manage: no no yes (ref-walk) maybe
52 ================== =========== ======== ============== ========
59 int (*create) (struct mnt_idmap *, struct inode *,struct dentry *,umode_t, bool);
60 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
61 int (*link) (struct dentry *,struct inode *,struct dentry *);
62 int (*unlink) (struct inode *,struct dentry *);
63 int (*symlink) (struct mnt_idmap *, struct inode *,struct dentry *,const char *);
64 int (*mkdir) (struct mnt_idmap *, struct inode *,struct dentry *,umode_t);
65 int (*rmdir) (struct inode *,struct dentry *);
66 int (*mknod) (struct mnt_idmap *, struct inode *,struct dentry *,umode_t,dev_t);
67 int (*rename) (struct mnt_idmap *, struct inode *, struct dentry *,
68 struct inode *, struct dentry *, unsigned int);
69 int (*readlink) (struct dentry *, char __user *,int);
70 const char *(*get_link) (struct dentry *, struct inode *, struct delayed_call *);
71 void (*truncate) (struct inode *);
72 int (*permission) (struct mnt_idmap *, struct inode *, int, unsigned int);
73 struct posix_acl * (*get_inode_acl)(struct inode *, int, bool);
74 int (*setattr) (struct mnt_idmap *, struct dentry *, struct iattr *);
75 int (*getattr) (struct mnt_idmap *, const struct path *, struct kstat *, u32, unsigned int);
76 ssize_t (*listxattr) (struct dentry *, char *, size_t);
77 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start, u64 len);
78 void (*update_time)(struct inode *, struct timespec *, int);
79 int (*atomic_open)(struct inode *, struct dentry *,
80 struct file *, unsigned open_flag,
82 int (*tmpfile) (struct mnt_idmap *, struct inode *,
83 struct file *, umode_t);
84 int (*fileattr_set)(struct mnt_idmap *idmap,
85 struct dentry *dentry, struct fileattr *fa);
86 int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa);
87 struct posix_acl * (*get_acl)(struct mnt_idmap *, struct dentry *, int);
88 struct offset_ctx *(*get_offset_ctx)(struct inode *inode);
93 ============== ==================================================
95 ============== ==================================================
98 link: exclusive (both)
102 unlink: exclusive (both)
103 rmdir: exclusive (both)(see below)
104 rename: exclusive (all) (see below)
108 permission: no (may not block if called in rcu-walk mode)
115 atomic_open: shared (exclusive if O_CREAT is set in open flags)
117 fileattr_get: no or exclusive
118 fileattr_set: exclusive
120 ============== ==================================================
123 Additionally, ->rmdir(), ->unlink() and ->rename() have ->i_rwsem
125 cross-directory ->rename() has (per-superblock) ->s_vfs_rename_sem.
127 See Documentation/filesystems/directory-locking.rst for more detailed discussion
128 of the locking scheme for directory operations.
130 xattr_handler operations
131 ========================
135 bool (*list)(struct dentry *dentry);
136 int (*get)(const struct xattr_handler *handler, struct dentry *dentry,
137 struct inode *inode, const char *name, void *buffer,
139 int (*set)(const struct xattr_handler *handler,
140 struct mnt_idmap *idmap,
141 struct dentry *dentry, struct inode *inode, const char *name,
142 const void *buffer, size_t size, int flags);
160 struct inode *(*alloc_inode)(struct super_block *sb);
161 void (*free_inode)(struct inode *);
162 void (*destroy_inode)(struct inode *);
163 void (*dirty_inode) (struct inode *, int flags);
164 int (*write_inode) (struct inode *, struct writeback_control *wbc);
165 int (*drop_inode) (struct inode *);
166 void (*evict_inode) (struct inode *);
167 void (*put_super) (struct super_block *);
168 int (*sync_fs)(struct super_block *sb, int wait);
169 int (*freeze_fs) (struct super_block *);
170 int (*unfreeze_fs) (struct super_block *);
171 int (*statfs) (struct dentry *, struct kstatfs *);
172 int (*remount_fs) (struct super_block *, int *, char *);
173 void (*umount_begin) (struct super_block *);
174 int (*show_options)(struct seq_file *, struct dentry *);
175 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
176 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
179 All may block [not true, see below]
181 ====================== ============ ========================
183 ====================== ============ ========================
185 free_inode: called from RCU callback
189 drop_inode: !!!inode->i_lock!!!
195 statfs: maybe(read) (see below)
198 show_options: no (namespace_sem)
199 quota_read: no (see below)
200 quota_write: no (see below)
201 ====================== ============ ========================
203 ->statfs() has s_umount (shared) when called by ustat(2) (native or
204 compat), but that's an accident of bad API; s_umount is used to pin
205 the superblock down when we only have dev_t given us by userland to
206 identify the superblock. Everything else (statfs(), fstatfs(), etc.)
207 doesn't hold it when calling ->statfs() - superblock is pinned down
208 by resolving the pathname passed to syscall.
210 ->quota_read() and ->quota_write() functions are both guaranteed to
211 be the only ones operating on the quota file by the quota code (via
212 dqio_sem) (unless an admin really wants to screw up something and
213 writes to quota files with quotas on). For other details about locking
214 see also dquot_operations section.
221 struct dentry *(*mount) (struct file_system_type *, int,
222 const char *, void *);
223 void (*kill_sb) (struct super_block *);
234 ->mount() returns ERR_PTR or the root dentry; its superblock should be locked
237 ->kill_sb() takes a write-locked superblock, does all shutdown work on it,
238 unlocks and drops the reference.
240 address_space_operations
241 ========================
244 int (*writepage)(struct page *page, struct writeback_control *wbc);
245 int (*read_folio)(struct file *, struct folio *);
246 int (*writepages)(struct address_space *, struct writeback_control *);
247 bool (*dirty_folio)(struct address_space *, struct folio *folio);
248 void (*readahead)(struct readahead_control *);
249 int (*write_begin)(struct file *, struct address_space *mapping,
250 loff_t pos, unsigned len,
251 struct page **pagep, void **fsdata);
252 int (*write_end)(struct file *, struct address_space *mapping,
253 loff_t pos, unsigned len, unsigned copied,
254 struct page *page, void *fsdata);
255 sector_t (*bmap)(struct address_space *, sector_t);
256 void (*invalidate_folio) (struct folio *, size_t start, size_t len);
257 bool (*release_folio)(struct folio *, gfp_t);
258 void (*free_folio)(struct folio *);
259 int (*direct_IO)(struct kiocb *, struct iov_iter *iter);
260 int (*migrate_folio)(struct address_space *, struct folio *dst,
261 struct folio *src, enum migrate_mode);
262 int (*launder_folio)(struct folio *);
263 bool (*is_partially_uptodate)(struct folio *, size_t from, size_t count);
264 int (*error_remove_page)(struct address_space *, struct page *);
265 int (*swap_activate)(struct swap_info_struct *sis, struct file *f, sector_t *span)
266 int (*swap_deactivate)(struct file *);
267 int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter);
270 All except dirty_folio and free_folio may block
272 ====================== ======================== ========= ===============
273 ops folio locked i_rwsem invalidate_lock
274 ====================== ======================== ========= ===============
275 writepage: yes, unlocks (see below)
276 read_folio: yes, unlocks shared
279 readahead: yes, unlocks shared
280 write_begin: locks the page exclusive
281 write_end: yes, unlocks exclusive
283 invalidate_folio: yes exclusive
287 migrate_folio: yes (both)
289 is_partially_uptodate: yes
290 error_remove_page: yes
293 swap_rw: yes, unlocks
294 ====================== ======================== ========= ===============
296 ->write_begin(), ->write_end() and ->read_folio() may be called from
297 the request handler (/dev/loop).
299 ->read_folio() unlocks the folio, either synchronously or via I/O
302 ->readahead() unlocks the folios that I/O is attempted on like ->read_folio().
304 ->writepage() is used for two purposes: for "memory cleansing" and for
305 "sync". These are quite different operations and the behaviour may differ
306 depending upon the mode.
308 If writepage is called for sync (wbc->sync_mode != WBC_SYNC_NONE) then
309 it *must* start I/O against the page, even if that would involve
310 blocking on in-progress I/O.
312 If writepage is called for memory cleansing (sync_mode ==
313 WBC_SYNC_NONE) then its role is to get as much writeout underway as
314 possible. So writepage should try to avoid blocking against
315 currently-in-progress I/O.
317 If the filesystem is not called for "sync" and it determines that it
318 would need to block against in-progress I/O to be able to start new I/O
319 against the page the filesystem should redirty the page with
320 redirty_page_for_writepage(), then unlock the page and return zero.
321 This may also be done to avoid internal deadlocks, but rarely.
323 If the filesystem is called for sync then it must wait on any
324 in-progress I/O and then start new I/O.
326 The filesystem should unlock the page synchronously, before returning to the
327 caller, unless ->writepage() returns special WRITEPAGE_ACTIVATE
328 value. WRITEPAGE_ACTIVATE means that page cannot really be written out
329 currently, and VM should stop calling ->writepage() on this page for some
330 time. VM does this by moving page to the head of the active list, hence the
333 Unless the filesystem is going to redirty_page_for_writepage(), unlock the page
334 and return zero, writepage *must* run set_page_writeback() against the page,
335 followed by unlocking it. Once set_page_writeback() has been run against the
336 page, write I/O can be submitted and the write I/O completion handler must run
337 end_page_writeback() once the I/O is complete. If no I/O is submitted, the
338 filesystem must run end_page_writeback() against the page before returning from
341 That is: after 2.5.12, pages which are under writeout are *not* locked. Note,
342 if the filesystem needs the page to be locked during writeout, that is ok, too,
343 the page is allowed to be unlocked at any point in time between the calls to
344 set_page_writeback() and end_page_writeback().
346 Note, failure to run either redirty_page_for_writepage() or the combination of
347 set_page_writeback()/end_page_writeback() on a page submitted to writepage
348 will leave the page itself marked clean but it will be tagged as dirty in the
349 radix tree. This incoherency can lead to all sorts of hard-to-debug problems
350 in the filesystem like having dirty inodes at umount and losing written data.
352 ->writepages() is used for periodic writeback and for syscall-initiated
353 sync operations. The address_space should start I/O against at least
354 ``*nr_to_write`` pages. ``*nr_to_write`` must be decremented for each page
355 which is written. The address_space implementation may write more (or less)
356 pages than ``*nr_to_write`` asks for, but it should try to be reasonably close.
357 If nr_to_write is NULL, all dirty pages must be written.
359 writepages should _only_ write pages which are present on
362 ->dirty_folio() is called from various places in the kernel when
363 the target folio is marked as needing writeback. The folio cannot be
364 truncated because either the caller holds the folio lock, or the caller
365 has found the folio while holding the page table lock which will block
368 ->bmap() is currently used by legacy ioctl() (FIBMAP) provided by some
369 filesystems and by the swapper. The latter will eventually go away. Please,
370 keep it that way and don't breed new callers.
372 ->invalidate_folio() is called when the filesystem must attempt to drop
373 some or all of the buffers from the page when it is being truncated. It
374 returns zero on success. The filesystem must exclusively acquire
375 invalidate_lock before invalidating page cache in truncate / hole punch
376 path (and thus calling into ->invalidate_folio) to block races between page
377 cache invalidation and page cache filling functions (fault, read, ...).
379 ->release_folio() is called when the MM wants to make a change to the
380 folio that would invalidate the filesystem's private data. For example,
381 it may be about to be removed from the address_space or split. The folio
382 is locked and not under writeback. It may be dirty. The gfp parameter
383 is not usually used for allocation, but rather to indicate what the
384 filesystem may do to attempt to free the private data. The filesystem may
385 return false to indicate that the folio's private data cannot be freed.
386 If it returns true, it should have already removed the private data from
387 the folio. If a filesystem does not provide a ->release_folio method,
388 the pagecache will assume that private data is buffer_heads and call
389 try_to_free_buffers().
391 ->free_folio() is called when the kernel has dropped the folio
394 ->launder_folio() may be called prior to releasing a folio if
395 it is still found to be dirty. It returns zero if the folio was successfully
396 cleaned, or an error value if not. Note that in order to prevent the folio
397 getting mapped back in and redirtied, it needs to be kept locked
398 across the entire operation.
400 ->swap_activate() will be called to prepare the given file for swap. It
401 should perform any validation and preparation necessary to ensure that
402 writes can be performed with minimal memory allocation. It should call
403 add_swap_extent(), or the helper iomap_swapfile_activate(), and return
404 the number of extents added. If IO should be submitted through
405 ->swap_rw(), it should set SWP_FS_OPS, otherwise IO will be submitted
406 directly to the block device ``sis->bdev``.
408 ->swap_deactivate() will be called in the sys_swapoff()
409 path after ->swap_activate() returned success.
411 ->swap_rw will be called for swap IO if SWP_FS_OPS was set by ->swap_activate().
418 void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
419 void (*fl_release_private)(struct file_lock *);
424 =================== ============= =========
425 ops inode->i_lock may block
426 =================== ============= =========
428 fl_release_private: maybe maybe[1]_
429 =================== ============= =========
432 ->fl_release_private for flock or POSIX locks is currently allowed
433 to block. Leases however can still be freed while the i_lock is held and
434 so fl_release_private called on a lease should not block.
436 lock_manager_operations
437 =======================
441 void (*lm_notify)(struct file_lock *); /* unblock callback */
442 int (*lm_grant)(struct file_lock *, struct file_lock *, int);
443 void (*lm_break)(struct file_lock *); /* break_lease callback */
444 int (*lm_change)(struct file_lock **, int);
445 bool (*lm_breaker_owns_lease)(struct file_lock *);
446 bool (*lm_lock_expirable)(struct file_lock *);
447 void (*lm_expire_lock)(void);
451 ====================== ============= ================= =========
452 ops flc_lock blocked_lock_lock may block
453 ====================== ============= ================= =========
458 lm_breaker_owns_lease: yes no no
459 lm_lock_expirable yes no no
460 lm_expire_lock no no yes
461 ====================== ============= ================= =========
468 void (*b_end_io)(struct buffer_head *bh, int uptodate);
472 called from interrupts. In other words, extreme care is needed here.
473 bh is locked, but that's all warranties we have here. Currently only RAID1,
474 highmem, fs/buffer.c, and fs/ntfs/aops.c are providing these. Block devices
475 call this method upon the IO completion.
477 block_device_operations
478 =======================
481 int (*open) (struct block_device *, fmode_t);
482 int (*release) (struct gendisk *, fmode_t);
483 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
484 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
485 int (*direct_access) (struct block_device *, sector_t, void **,
487 void (*unlock_native_capacity) (struct gendisk *);
488 int (*getgeo)(struct block_device *, struct hd_geometry *);
489 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
493 ======================= ===================
495 ======================= ===================
501 unlock_native_capacity: no
503 swap_slot_free_notify: no (see below)
504 ======================= ===================
506 swap_slot_free_notify is called with swap_lock and sometimes the page lock
515 loff_t (*llseek) (struct file *, loff_t, int);
516 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
517 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
518 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
519 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
520 int (*iopoll) (struct kiocb *kiocb, bool spin);
521 int (*iterate_shared) (struct file *, struct dir_context *);
522 __poll_t (*poll) (struct file *, struct poll_table_struct *);
523 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
524 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
525 int (*mmap) (struct file *, struct vm_area_struct *);
526 int (*open) (struct inode *, struct file *);
527 int (*flush) (struct file *);
528 int (*release) (struct inode *, struct file *);
529 int (*fsync) (struct file *, loff_t start, loff_t end, int datasync);
530 int (*fasync) (int, struct file *, int);
531 int (*lock) (struct file *, int, struct file_lock *);
532 unsigned long (*get_unmapped_area)(struct file *, unsigned long,
533 unsigned long, unsigned long, unsigned long);
534 int (*check_flags)(int);
535 int (*flock) (struct file *, int, struct file_lock *);
536 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *,
537 size_t, unsigned int);
538 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *,
539 size_t, unsigned int);
540 int (*setlease)(struct file *, long, struct file_lock **, void **);
541 long (*fallocate)(struct file *, int, loff_t, loff_t);
542 void (*show_fdinfo)(struct seq_file *m, struct file *f);
543 unsigned (*mmap_capabilities)(struct file *);
544 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *,
545 loff_t, size_t, unsigned int);
546 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in,
547 struct file *file_out, loff_t pos_out,
548 loff_t len, unsigned int remap_flags);
549 int (*fadvise)(struct file *, loff_t, loff_t, int);
554 ->llseek() locking has moved from llseek to the individual llseek
555 implementations. If your fs is not using generic_file_llseek, you
556 need to acquire and release the appropriate locks in your ->llseek().
557 For many filesystems, it is probably safe to acquire the inode
558 mutex or just to use i_size_read() instead.
559 Note: this does not protect the file->f_pos against concurrent modifications
560 since this is something the userspace has to take care about.
562 ->iterate_shared() is called with i_rwsem held for reading, and with the
563 file f_pos_lock held exclusively
565 ->fasync() is responsible for maintaining the FASYNC bit in filp->f_flags.
566 Most instances call fasync_helper(), which does that maintenance, so it's
567 not normally something one needs to worry about. Return values > 0 will be
568 mapped to zero in the VFS layer.
570 ->readdir() and ->ioctl() on directories must be changed. Ideally we would
571 move ->readdir() to inode_operations and use a separate method for directory
572 ->ioctl() or kill the latter completely. One of the problems is that for
573 anything that resembles union-mount we won't have a struct file for all
574 components. And there are other reasons why the current interface is a mess...
576 ->read on directories probably must go away - we should just enforce -EISDIR
577 in sys_read() and friends.
579 ->setlease operations should call generic_setlease() before or after setting
580 the lease within the individual filesystem to record the result of the
583 ->fallocate implementation must be really careful to maintain page cache
584 consistency when punching holes or performing other operations that invalidate
585 page cache contents. Usually the filesystem needs to call
586 truncate_inode_pages_range() to invalidate relevant range of the page cache.
587 However the filesystem usually also needs to update its internal (and on disk)
588 view of file offset -> disk block mapping. Until this update is finished, the
589 filesystem needs to block page faults and reads from reloading now-stale page
590 cache contents from the disk. Since VFS acquires mapping->invalidate_lock in
591 shared mode when loading pages from disk (filemap_fault(), filemap_read(),
592 readahead paths), the fallocate implementation must take the invalidate_lock to
595 ->copy_file_range and ->remap_file_range implementations need to serialize
596 against modifications of file data while the operation is running. For
597 blocking changes through write(2) and similar operations inode->i_rwsem can be
598 used. To block changes to file contents via a memory mapping during the
599 operation, the filesystem must take mapping->invalidate_lock to coordinate
607 int (*write_dquot) (struct dquot *);
608 int (*acquire_dquot) (struct dquot *);
609 int (*release_dquot) (struct dquot *);
610 int (*mark_dirty) (struct dquot *);
611 int (*write_info) (struct super_block *, int);
613 These operations are intended to be more or less wrapping functions that ensure
614 a proper locking wrt the filesystem and call the generic quota operations.
616 What filesystem should expect from the generic quota functions:
618 ============== ============ =========================
619 ops FS recursion Held locks when called
620 ============== ============ =========================
621 write_dquot: yes dqonoff_sem or dqptr_sem
622 acquire_dquot: yes dqonoff_sem or dqptr_sem
623 release_dquot: yes dqonoff_sem or dqptr_sem
625 write_info: yes dqonoff_sem
626 ============== ============ =========================
628 FS recursion means calling ->quota_read() and ->quota_write() from superblock
631 More details about quota locking can be found in fs/dquot.c.
638 void (*open)(struct vm_area_struct *);
639 void (*close)(struct vm_area_struct *);
640 vm_fault_t (*fault)(struct vm_fault *);
641 vm_fault_t (*huge_fault)(struct vm_fault *, unsigned int order);
642 vm_fault_t (*map_pages)(struct vm_fault *, pgoff_t start, pgoff_t end);
643 vm_fault_t (*page_mkwrite)(struct vm_area_struct *, struct vm_fault *);
644 vm_fault_t (*pfn_mkwrite)(struct vm_area_struct *, struct vm_fault *);
645 int (*access)(struct vm_area_struct *, unsigned long, void*, int, int);
649 ============= ========== ===========================
650 ops mmap_lock PageLocked(page)
651 ============= ========== ===========================
654 fault: read can return with page locked
655 huge_fault: maybe-read
656 map_pages: maybe-read
657 page_mkwrite: read can return with page locked
660 ============= ========== ===========================
662 ->fault() is called when a previously not present pte is about to be faulted
663 in. The filesystem must find and return the page associated with the passed in
664 "pgoff" in the vm_fault structure. If it is possible that the page may be
665 truncated and/or invalidated, then the filesystem must lock invalidate_lock,
666 then ensure the page is not already truncated (invalidate_lock will block
667 subsequent truncate), and then return with VM_FAULT_LOCKED, and the page
668 locked. The VM will unlock the page.
670 ->huge_fault() is called when there is no PUD or PMD entry present. This
671 gives the filesystem the opportunity to install a PUD or PMD sized page.
672 Filesystems can also use the ->fault method to return a PMD sized page,
673 so implementing this function may not be necessary. In particular,
674 filesystems should not call filemap_fault() from ->huge_fault().
675 The mmap_lock may not be held when this method is called.
677 ->map_pages() is called when VM asks to map easy accessible pages.
678 Filesystem should find and map pages associated with offsets from "start_pgoff"
679 till "end_pgoff". ->map_pages() is called with the RCU lock held and must
680 not block. If it's not possible to reach a page without blocking,
681 filesystem should skip it. Filesystem should use set_pte_range() to setup
682 page table entry. Pointer to entry associated with the page is passed in
683 "pte" field in vm_fault structure. Pointers to entries for other offsets
684 should be calculated relative to "pte".
686 ->page_mkwrite() is called when a previously read-only pte is about to become
687 writeable. The filesystem again must ensure that there are no
688 truncate/invalidate races or races with operations such as ->remap_file_range
689 or ->copy_file_range, and then return with the page locked. Usually
690 mapping->invalidate_lock is suitable for proper serialization. If the page has
691 been truncated, the filesystem should not look up a new page like the ->fault()
692 handler, but simply return with VM_FAULT_NOPAGE, which will cause the VM to
695 ->pfn_mkwrite() is the same as page_mkwrite but when the pte is
696 VM_PFNMAP or VM_MIXEDMAP with a page-less entry. Expected return is
697 VM_FAULT_NOPAGE. Or one of the VM_FAULT_ERROR types. The default behavior
698 after this call is to make the pte read-write, unless pfn_mkwrite returns
701 ->access() is called when get_user_pages() fails in
702 access_process_vm(), typically used to debug a process through
703 /proc/pid/mem or ptrace. This function is needed only for
704 VM_IO | VM_PFNMAP VMAs.
706 --------------------------------------------------------------------------------
710 (if you break something or notice that it is broken and do not fix it yourself
711 - at least put it here)