Loading...
1=======
2Locking
3=======
4
5The text below describes the locking rules for VFS-related methods.
6It is (believed to be) up-to-date. *Please*, if you change anything in
7prototypes or locking protocols - update this file. And update the relevant
8instances in the tree, don't leave that to maintainers of filesystems/devices/
9etc. At the very least, put the list of dubious cases in the end of this file.
10Don't turn it into log - maintainers of out-of-the-tree code are supposed to
11be able to use diff(1).
12
13Thing currently missing here: socket operations. Alexey?
14
15dentry_operations
16=================
17
18prototypes::
19
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 *);
33
34locking rules:
35
36================== =========== ======== ============== ========
37ops rename_lock ->d_lock may block rcu-walk
38================== =========== ======== ============== ========
39d_revalidate: no no yes (ref-walk) maybe
40d_weak_revalidate: no no yes no
41d_hash no no no maybe
42d_compare: yes no no maybe
43d_delete: no yes no no
44d_init: no no yes no
45d_release: no no yes no
46d_prune: no yes no no
47d_iput: no no yes no
48d_dname: no no no no
49d_automount: no no yes no
50d_manage: no no yes (ref-walk) maybe
51d_real no no yes no
52================== =========== ======== ============== ========
53
54inode_operations
55================
56
57prototypes::
58
59 int (*create) (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 inode *,struct dentry *,const char *);
64 int (*mkdir) (struct inode *,struct dentry *,umode_t);
65 int (*rmdir) (struct inode *,struct dentry *);
66 int (*mknod) (struct inode *,struct dentry *,umode_t,dev_t);
67 int (*rename) (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 inode *, int, unsigned int);
73 int (*get_acl)(struct inode *, int);
74 int (*setattr) (struct dentry *, struct iattr *);
75 int (*getattr) (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,
81 umode_t create_mode);
82 int (*tmpfile) (struct inode *, struct dentry *, umode_t);
83 int (*fileattr_set)(struct user_namespace *mnt_userns,
84 struct dentry *dentry, struct fileattr *fa);
85 int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa);
86
87locking rules:
88 all may block
89
90============= =============================================
91ops i_rwsem(inode)
92============= =============================================
93lookup: shared
94create: exclusive
95link: exclusive (both)
96mknod: exclusive
97symlink: exclusive
98mkdir: exclusive
99unlink: exclusive (both)
100rmdir: exclusive (both)(see below)
101rename: exclusive (all) (see below)
102readlink: no
103get_link: no
104setattr: exclusive
105permission: no (may not block if called in rcu-walk mode)
106get_acl: no
107getattr: no
108listxattr: no
109fiemap: no
110update_time: no
111atomic_open: shared (exclusive if O_CREAT is set in open flags)
112tmpfile: no
113fileattr_get: no or exclusive
114fileattr_set: exclusive
115============= =============================================
116
117
118 Additionally, ->rmdir(), ->unlink() and ->rename() have ->i_rwsem
119 exclusive on victim.
120 cross-directory ->rename() has (per-superblock) ->s_vfs_rename_sem.
121
122See Documentation/filesystems/directory-locking.rst for more detailed discussion
123of the locking scheme for directory operations.
124
125xattr_handler operations
126========================
127
128prototypes::
129
130 bool (*list)(struct dentry *dentry);
131 int (*get)(const struct xattr_handler *handler, struct dentry *dentry,
132 struct inode *inode, const char *name, void *buffer,
133 size_t size);
134 int (*set)(const struct xattr_handler *handler,
135 struct user_namespace *mnt_userns,
136 struct dentry *dentry, struct inode *inode, const char *name,
137 const void *buffer, size_t size, int flags);
138
139locking rules:
140 all may block
141
142===== ==============
143ops i_rwsem(inode)
144===== ==============
145list: no
146get: no
147set: exclusive
148===== ==============
149
150super_operations
151================
152
153prototypes::
154
155 struct inode *(*alloc_inode)(struct super_block *sb);
156 void (*free_inode)(struct inode *);
157 void (*destroy_inode)(struct inode *);
158 void (*dirty_inode) (struct inode *, int flags);
159 int (*write_inode) (struct inode *, struct writeback_control *wbc);
160 int (*drop_inode) (struct inode *);
161 void (*evict_inode) (struct inode *);
162 void (*put_super) (struct super_block *);
163 int (*sync_fs)(struct super_block *sb, int wait);
164 int (*freeze_fs) (struct super_block *);
165 int (*unfreeze_fs) (struct super_block *);
166 int (*statfs) (struct dentry *, struct kstatfs *);
167 int (*remount_fs) (struct super_block *, int *, char *);
168 void (*umount_begin) (struct super_block *);
169 int (*show_options)(struct seq_file *, struct dentry *);
170 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
171 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
172 int (*bdev_try_to_free_page)(struct super_block*, struct page*, gfp_t);
173
174locking rules:
175 All may block [not true, see below]
176
177====================== ============ ========================
178ops s_umount note
179====================== ============ ========================
180alloc_inode:
181free_inode: called from RCU callback
182destroy_inode:
183dirty_inode:
184write_inode:
185drop_inode: !!!inode->i_lock!!!
186evict_inode:
187put_super: write
188sync_fs: read
189freeze_fs: write
190unfreeze_fs: write
191statfs: maybe(read) (see below)
192remount_fs: write
193umount_begin: no
194show_options: no (namespace_sem)
195quota_read: no (see below)
196quota_write: no (see below)
197bdev_try_to_free_page: no (see below)
198====================== ============ ========================
199
200->statfs() has s_umount (shared) when called by ustat(2) (native or
201compat), but that's an accident of bad API; s_umount is used to pin
202the superblock down when we only have dev_t given us by userland to
203identify the superblock. Everything else (statfs(), fstatfs(), etc.)
204doesn't hold it when calling ->statfs() - superblock is pinned down
205by resolving the pathname passed to syscall.
206
207->quota_read() and ->quota_write() functions are both guaranteed to
208be the only ones operating on the quota file by the quota code (via
209dqio_sem) (unless an admin really wants to screw up something and
210writes to quota files with quotas on). For other details about locking
211see also dquot_operations section.
212
213->bdev_try_to_free_page is called from the ->releasepage handler of
214the block device inode. See there for more details.
215
216file_system_type
217================
218
219prototypes::
220
221 struct dentry *(*mount) (struct file_system_type *, int,
222 const char *, void *);
223 void (*kill_sb) (struct super_block *);
224
225locking rules:
226
227======= =========
228ops may block
229======= =========
230mount yes
231kill_sb yes
232======= =========
233
234->mount() returns ERR_PTR or the root dentry; its superblock should be locked
235on return.
236
237->kill_sb() takes a write-locked superblock, does all shutdown work on it,
238unlocks and drops the reference.
239
240address_space_operations
241========================
242prototypes::
243
244 int (*writepage)(struct page *page, struct writeback_control *wbc);
245 int (*readpage)(struct file *, struct page *);
246 int (*writepages)(struct address_space *, struct writeback_control *);
247 int (*set_page_dirty)(struct page *page);
248 void (*readahead)(struct readahead_control *);
249 int (*readpages)(struct file *filp, struct address_space *mapping,
250 struct list_head *pages, unsigned nr_pages);
251 int (*write_begin)(struct file *, struct address_space *mapping,
252 loff_t pos, unsigned len, unsigned flags,
253 struct page **pagep, void **fsdata);
254 int (*write_end)(struct file *, struct address_space *mapping,
255 loff_t pos, unsigned len, unsigned copied,
256 struct page *page, void *fsdata);
257 sector_t (*bmap)(struct address_space *, sector_t);
258 void (*invalidatepage) (struct page *, unsigned int, unsigned int);
259 int (*releasepage) (struct page *, int);
260 void (*freepage)(struct page *);
261 int (*direct_IO)(struct kiocb *, struct iov_iter *iter);
262 bool (*isolate_page) (struct page *, isolate_mode_t);
263 int (*migratepage)(struct address_space *, struct page *, struct page *);
264 void (*putback_page) (struct page *);
265 int (*launder_page)(struct page *);
266 int (*is_partially_uptodate)(struct page *, unsigned long, unsigned long);
267 int (*error_remove_page)(struct address_space *, struct page *);
268 int (*swap_activate)(struct file *);
269 int (*swap_deactivate)(struct file *);
270
271locking rules:
272 All except set_page_dirty and freepage may block
273
274====================== ======================== =========
275ops PageLocked(page) i_rwsem
276====================== ======================== =========
277writepage: yes, unlocks (see below)
278readpage: yes, unlocks
279writepages:
280set_page_dirty no
281readahead: yes, unlocks
282readpages: no
283write_begin: locks the page exclusive
284write_end: yes, unlocks exclusive
285bmap:
286invalidatepage: yes
287releasepage: yes
288freepage: yes
289direct_IO:
290isolate_page: yes
291migratepage: yes (both)
292putback_page: yes
293launder_page: yes
294is_partially_uptodate: yes
295error_remove_page: yes
296swap_activate: no
297swap_deactivate: no
298====================== ======================== =========
299
300->write_begin(), ->write_end() and ->readpage() may be called from
301the request handler (/dev/loop).
302
303->readpage() unlocks the page, either synchronously or via I/O
304completion.
305
306->readahead() unlocks the pages that I/O is attempted on like ->readpage().
307
308->readpages() populates the pagecache with the passed pages and starts
309I/O against them. They come unlocked upon I/O completion.
310
311->writepage() is used for two purposes: for "memory cleansing" and for
312"sync". These are quite different operations and the behaviour may differ
313depending upon the mode.
314
315If writepage is called for sync (wbc->sync_mode != WBC_SYNC_NONE) then
316it *must* start I/O against the page, even if that would involve
317blocking on in-progress I/O.
318
319If writepage is called for memory cleansing (sync_mode ==
320WBC_SYNC_NONE) then its role is to get as much writeout underway as
321possible. So writepage should try to avoid blocking against
322currently-in-progress I/O.
323
324If the filesystem is not called for "sync" and it determines that it
325would need to block against in-progress I/O to be able to start new I/O
326against the page the filesystem should redirty the page with
327redirty_page_for_writepage(), then unlock the page and return zero.
328This may also be done to avoid internal deadlocks, but rarely.
329
330If the filesystem is called for sync then it must wait on any
331in-progress I/O and then start new I/O.
332
333The filesystem should unlock the page synchronously, before returning to the
334caller, unless ->writepage() returns special WRITEPAGE_ACTIVATE
335value. WRITEPAGE_ACTIVATE means that page cannot really be written out
336currently, and VM should stop calling ->writepage() on this page for some
337time. VM does this by moving page to the head of the active list, hence the
338name.
339
340Unless the filesystem is going to redirty_page_for_writepage(), unlock the page
341and return zero, writepage *must* run set_page_writeback() against the page,
342followed by unlocking it. Once set_page_writeback() has been run against the
343page, write I/O can be submitted and the write I/O completion handler must run
344end_page_writeback() once the I/O is complete. If no I/O is submitted, the
345filesystem must run end_page_writeback() against the page before returning from
346writepage.
347
348That is: after 2.5.12, pages which are under writeout are *not* locked. Note,
349if the filesystem needs the page to be locked during writeout, that is ok, too,
350the page is allowed to be unlocked at any point in time between the calls to
351set_page_writeback() and end_page_writeback().
352
353Note, failure to run either redirty_page_for_writepage() or the combination of
354set_page_writeback()/end_page_writeback() on a page submitted to writepage
355will leave the page itself marked clean but it will be tagged as dirty in the
356radix tree. This incoherency can lead to all sorts of hard-to-debug problems
357in the filesystem like having dirty inodes at umount and losing written data.
358
359->writepages() is used for periodic writeback and for syscall-initiated
360sync operations. The address_space should start I/O against at least
361``*nr_to_write`` pages. ``*nr_to_write`` must be decremented for each page
362which is written. The address_space implementation may write more (or less)
363pages than ``*nr_to_write`` asks for, but it should try to be reasonably close.
364If nr_to_write is NULL, all dirty pages must be written.
365
366writepages should _only_ write pages which are present on
367mapping->io_pages.
368
369->set_page_dirty() is called from various places in the kernel
370when the target page is marked as needing writeback. It may be called
371under spinlock (it cannot block) and is sometimes called with the page
372not locked.
373
374->bmap() is currently used by legacy ioctl() (FIBMAP) provided by some
375filesystems and by the swapper. The latter will eventually go away. Please,
376keep it that way and don't breed new callers.
377
378->invalidatepage() is called when the filesystem must attempt to drop
379some or all of the buffers from the page when it is being truncated. It
380returns zero on success. If ->invalidatepage is zero, the kernel uses
381block_invalidatepage() instead.
382
383->releasepage() is called when the kernel is about to try to drop the
384buffers from the page in preparation for freeing it. It returns zero to
385indicate that the buffers are (or may be) freeable. If ->releasepage is zero,
386the kernel assumes that the fs has no private interest in the buffers.
387
388->freepage() is called when the kernel is done dropping the page
389from the page cache.
390
391->launder_page() may be called prior to releasing a page if
392it is still found to be dirty. It returns zero if the page was successfully
393cleaned, or an error value if not. Note that in order to prevent the page
394getting mapped back in and redirtied, it needs to be kept locked
395across the entire operation.
396
397->swap_activate will be called with a non-zero argument on
398files backing (non block device backed) swapfiles. A return value
399of zero indicates success, in which case this file can be used for
400backing swapspace. The swapspace operations will be proxied to the
401address space operations.
402
403->swap_deactivate() will be called in the sys_swapoff()
404path after ->swap_activate() returned success.
405
406file_lock_operations
407====================
408
409prototypes::
410
411 void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
412 void (*fl_release_private)(struct file_lock *);
413
414
415locking rules:
416
417=================== ============= =========
418ops inode->i_lock may block
419=================== ============= =========
420fl_copy_lock: yes no
421fl_release_private: maybe maybe[1]_
422=================== ============= =========
423
424.. [1]:
425 ->fl_release_private for flock or POSIX locks is currently allowed
426 to block. Leases however can still be freed while the i_lock is held and
427 so fl_release_private called on a lease should not block.
428
429lock_manager_operations
430=======================
431
432prototypes::
433
434 void (*lm_notify)(struct file_lock *); /* unblock callback */
435 int (*lm_grant)(struct file_lock *, struct file_lock *, int);
436 void (*lm_break)(struct file_lock *); /* break_lease callback */
437 int (*lm_change)(struct file_lock **, int);
438 bool (*lm_breaker_owns_lease)(struct file_lock *);
439
440locking rules:
441
442====================== ============= ================= =========
443ops inode->i_lock blocked_lock_lock may block
444====================== ============= ================= =========
445lm_notify: yes yes no
446lm_grant: no no no
447lm_break: yes no no
448lm_change yes no no
449lm_breaker_owns_lease: no no no
450====================== ============= ================= =========
451
452buffer_head
453===========
454
455prototypes::
456
457 void (*b_end_io)(struct buffer_head *bh, int uptodate);
458
459locking rules:
460
461called from interrupts. In other words, extreme care is needed here.
462bh is locked, but that's all warranties we have here. Currently only RAID1,
463highmem, fs/buffer.c, and fs/ntfs/aops.c are providing these. Block devices
464call this method upon the IO completion.
465
466block_device_operations
467=======================
468prototypes::
469
470 int (*open) (struct block_device *, fmode_t);
471 int (*release) (struct gendisk *, fmode_t);
472 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
473 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
474 int (*direct_access) (struct block_device *, sector_t, void **,
475 unsigned long *);
476 void (*unlock_native_capacity) (struct gendisk *);
477 int (*getgeo)(struct block_device *, struct hd_geometry *);
478 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
479
480locking rules:
481
482======================= ===================
483ops open_mutex
484======================= ===================
485open: yes
486release: yes
487ioctl: no
488compat_ioctl: no
489direct_access: no
490unlock_native_capacity: no
491getgeo: no
492swap_slot_free_notify: no (see below)
493======================= ===================
494
495swap_slot_free_notify is called with swap_lock and sometimes the page lock
496held.
497
498
499file_operations
500===============
501
502prototypes::
503
504 loff_t (*llseek) (struct file *, loff_t, int);
505 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
506 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
507 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
508 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
509 int (*iterate) (struct file *, struct dir_context *);
510 int (*iterate_shared) (struct file *, struct dir_context *);
511 __poll_t (*poll) (struct file *, struct poll_table_struct *);
512 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
513 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
514 int (*mmap) (struct file *, struct vm_area_struct *);
515 int (*open) (struct inode *, struct file *);
516 int (*flush) (struct file *);
517 int (*release) (struct inode *, struct file *);
518 int (*fsync) (struct file *, loff_t start, loff_t end, int datasync);
519 int (*fasync) (int, struct file *, int);
520 int (*lock) (struct file *, int, struct file_lock *);
521 ssize_t (*readv) (struct file *, const struct iovec *, unsigned long,
522 loff_t *);
523 ssize_t (*writev) (struct file *, const struct iovec *, unsigned long,
524 loff_t *);
525 ssize_t (*sendfile) (struct file *, loff_t *, size_t, read_actor_t,
526 void __user *);
527 ssize_t (*sendpage) (struct file *, struct page *, int, size_t,
528 loff_t *, int);
529 unsigned long (*get_unmapped_area)(struct file *, unsigned long,
530 unsigned long, unsigned long, unsigned long);
531 int (*check_flags)(int);
532 int (*flock) (struct file *, int, struct file_lock *);
533 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *,
534 size_t, unsigned int);
535 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *,
536 size_t, unsigned int);
537 int (*setlease)(struct file *, long, struct file_lock **, void **);
538 long (*fallocate)(struct file *, int, loff_t, loff_t);
539
540locking rules:
541 All may block.
542
543->llseek() locking has moved from llseek to the individual llseek
544implementations. If your fs is not using generic_file_llseek, you
545need to acquire and release the appropriate locks in your ->llseek().
546For many filesystems, it is probably safe to acquire the inode
547mutex or just to use i_size_read() instead.
548Note: this does not protect the file->f_pos against concurrent modifications
549since this is something the userspace has to take care about.
550
551->iterate() is called with i_rwsem exclusive.
552
553->iterate_shared() is called with i_rwsem at least shared.
554
555->fasync() is responsible for maintaining the FASYNC bit in filp->f_flags.
556Most instances call fasync_helper(), which does that maintenance, so it's
557not normally something one needs to worry about. Return values > 0 will be
558mapped to zero in the VFS layer.
559
560->readdir() and ->ioctl() on directories must be changed. Ideally we would
561move ->readdir() to inode_operations and use a separate method for directory
562->ioctl() or kill the latter completely. One of the problems is that for
563anything that resembles union-mount we won't have a struct file for all
564components. And there are other reasons why the current interface is a mess...
565
566->read on directories probably must go away - we should just enforce -EISDIR
567in sys_read() and friends.
568
569->setlease operations should call generic_setlease() before or after setting
570the lease within the individual filesystem to record the result of the
571operation
572
573dquot_operations
574================
575
576prototypes::
577
578 int (*write_dquot) (struct dquot *);
579 int (*acquire_dquot) (struct dquot *);
580 int (*release_dquot) (struct dquot *);
581 int (*mark_dirty) (struct dquot *);
582 int (*write_info) (struct super_block *, int);
583
584These operations are intended to be more or less wrapping functions that ensure
585a proper locking wrt the filesystem and call the generic quota operations.
586
587What filesystem should expect from the generic quota functions:
588
589============== ============ =========================
590ops FS recursion Held locks when called
591============== ============ =========================
592write_dquot: yes dqonoff_sem or dqptr_sem
593acquire_dquot: yes dqonoff_sem or dqptr_sem
594release_dquot: yes dqonoff_sem or dqptr_sem
595mark_dirty: no -
596write_info: yes dqonoff_sem
597============== ============ =========================
598
599FS recursion means calling ->quota_read() and ->quota_write() from superblock
600operations.
601
602More details about quota locking can be found in fs/dquot.c.
603
604vm_operations_struct
605====================
606
607prototypes::
608
609 void (*open)(struct vm_area_struct*);
610 void (*close)(struct vm_area_struct*);
611 vm_fault_t (*fault)(struct vm_area_struct*, struct vm_fault *);
612 vm_fault_t (*page_mkwrite)(struct vm_area_struct *, struct vm_fault *);
613 vm_fault_t (*pfn_mkwrite)(struct vm_area_struct *, struct vm_fault *);
614 int (*access)(struct vm_area_struct *, unsigned long, void*, int, int);
615
616locking rules:
617
618============= ========= ===========================
619ops mmap_lock PageLocked(page)
620============= ========= ===========================
621open: yes
622close: yes
623fault: yes can return with page locked
624map_pages: yes
625page_mkwrite: yes can return with page locked
626pfn_mkwrite: yes
627access: yes
628============= ========= ===========================
629
630->fault() is called when a previously not present pte is about
631to be faulted in. The filesystem must find and return the page associated
632with the passed in "pgoff" in the vm_fault structure. If it is possible that
633the page may be truncated and/or invalidated, then the filesystem must lock
634the page, then ensure it is not already truncated (the page lock will block
635subsequent truncate), and then return with VM_FAULT_LOCKED, and the page
636locked. The VM will unlock the page.
637
638->map_pages() is called when VM asks to map easy accessible pages.
639Filesystem should find and map pages associated with offsets from "start_pgoff"
640till "end_pgoff". ->map_pages() is called with page table locked and must
641not block. If it's not possible to reach a page without blocking,
642filesystem should skip it. Filesystem should use do_set_pte() to setup
643page table entry. Pointer to entry associated with the page is passed in
644"pte" field in vm_fault structure. Pointers to entries for other offsets
645should be calculated relative to "pte".
646
647->page_mkwrite() is called when a previously read-only pte is
648about to become writeable. The filesystem again must ensure that there are
649no truncate/invalidate races, and then return with the page locked. If
650the page has been truncated, the filesystem should not look up a new page
651like the ->fault() handler, but simply return with VM_FAULT_NOPAGE, which
652will cause the VM to retry the fault.
653
654->pfn_mkwrite() is the same as page_mkwrite but when the pte is
655VM_PFNMAP or VM_MIXEDMAP with a page-less entry. Expected return is
656VM_FAULT_NOPAGE. Or one of the VM_FAULT_ERROR types. The default behavior
657after this call is to make the pte read-write, unless pfn_mkwrite returns
658an error.
659
660->access() is called when get_user_pages() fails in
661access_process_vm(), typically used to debug a process through
662/proc/pid/mem or ptrace. This function is needed only for
663VM_IO | VM_PFNMAP VMAs.
664
665--------------------------------------------------------------------------------
666
667 Dubious stuff
668
669(if you break something or notice that it is broken and do not fix it yourself
670- at least put it here)
1=======
2Locking
3=======
4
5The text below describes the locking rules for VFS-related methods.
6It is (believed to be) up-to-date. *Please*, if you change anything in
7prototypes or locking protocols - update this file. And update the relevant
8instances in the tree, don't leave that to maintainers of filesystems/devices/
9etc. At the very least, put the list of dubious cases in the end of this file.
10Don't turn it into log - maintainers of out-of-the-tree code are supposed to
11be able to use diff(1).
12
13Thing currently missing here: socket operations. Alexey?
14
15dentry_operations
16=================
17
18prototypes::
19
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 *);
33
34locking rules:
35
36================== =========== ======== ============== ========
37ops rename_lock ->d_lock may block rcu-walk
38================== =========== ======== ============== ========
39d_revalidate: no no yes (ref-walk) maybe
40d_weak_revalidate: no no yes no
41d_hash no no no maybe
42d_compare: yes no no maybe
43d_delete: no yes no no
44d_init: no no yes no
45d_release: no no yes no
46d_prune: no yes no no
47d_iput: no no yes no
48d_dname: no no no no
49d_automount: no no yes no
50d_manage: no no yes (ref-walk) maybe
51d_real no no yes no
52================== =========== ======== ============== ========
53
54inode_operations
55================
56
57prototypes::
58
59 int (*create) (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 inode *,struct dentry *,const char *);
64 int (*mkdir) (struct inode *,struct dentry *,umode_t);
65 int (*rmdir) (struct inode *,struct dentry *);
66 int (*mknod) (struct inode *,struct dentry *,umode_t,dev_t);
67 int (*rename) (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 inode *, int, unsigned int);
73 int (*get_acl)(struct inode *, int);
74 int (*setattr) (struct dentry *, struct iattr *);
75 int (*getattr) (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,
81 umode_t create_mode);
82 int (*tmpfile) (struct inode *, struct dentry *, umode_t);
83
84locking rules:
85 all may block
86
87============ =============================================
88ops i_rwsem(inode)
89============ =============================================
90lookup: shared
91create: exclusive
92link: exclusive (both)
93mknod: exclusive
94symlink: exclusive
95mkdir: exclusive
96unlink: exclusive (both)
97rmdir: exclusive (both)(see below)
98rename: exclusive (all) (see below)
99readlink: no
100get_link: no
101setattr: exclusive
102permission: no (may not block if called in rcu-walk mode)
103get_acl: no
104getattr: no
105listxattr: no
106fiemap: no
107update_time: no
108atomic_open: exclusive
109tmpfile: no
110============ =============================================
111
112
113 Additionally, ->rmdir(), ->unlink() and ->rename() have ->i_rwsem
114 exclusive on victim.
115 cross-directory ->rename() has (per-superblock) ->s_vfs_rename_sem.
116
117See Documentation/filesystems/directory-locking.rst for more detailed discussion
118of the locking scheme for directory operations.
119
120xattr_handler operations
121========================
122
123prototypes::
124
125 bool (*list)(struct dentry *dentry);
126 int (*get)(const struct xattr_handler *handler, struct dentry *dentry,
127 struct inode *inode, const char *name, void *buffer,
128 size_t size);
129 int (*set)(const struct xattr_handler *handler, struct dentry *dentry,
130 struct inode *inode, const char *name, const void *buffer,
131 size_t size, int flags);
132
133locking rules:
134 all may block
135
136===== ==============
137ops i_rwsem(inode)
138===== ==============
139list: no
140get: no
141set: exclusive
142===== ==============
143
144super_operations
145================
146
147prototypes::
148
149 struct inode *(*alloc_inode)(struct super_block *sb);
150 void (*free_inode)(struct inode *);
151 void (*destroy_inode)(struct inode *);
152 void (*dirty_inode) (struct inode *, int flags);
153 int (*write_inode) (struct inode *, struct writeback_control *wbc);
154 int (*drop_inode) (struct inode *);
155 void (*evict_inode) (struct inode *);
156 void (*put_super) (struct super_block *);
157 int (*sync_fs)(struct super_block *sb, int wait);
158 int (*freeze_fs) (struct super_block *);
159 int (*unfreeze_fs) (struct super_block *);
160 int (*statfs) (struct dentry *, struct kstatfs *);
161 int (*remount_fs) (struct super_block *, int *, char *);
162 void (*umount_begin) (struct super_block *);
163 int (*show_options)(struct seq_file *, struct dentry *);
164 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
165 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
166 int (*bdev_try_to_free_page)(struct super_block*, struct page*, gfp_t);
167
168locking rules:
169 All may block [not true, see below]
170
171====================== ============ ========================
172ops s_umount note
173====================== ============ ========================
174alloc_inode:
175free_inode: called from RCU callback
176destroy_inode:
177dirty_inode:
178write_inode:
179drop_inode: !!!inode->i_lock!!!
180evict_inode:
181put_super: write
182sync_fs: read
183freeze_fs: write
184unfreeze_fs: write
185statfs: maybe(read) (see below)
186remount_fs: write
187umount_begin: no
188show_options: no (namespace_sem)
189quota_read: no (see below)
190quota_write: no (see below)
191bdev_try_to_free_page: no (see below)
192====================== ============ ========================
193
194->statfs() has s_umount (shared) when called by ustat(2) (native or
195compat), but that's an accident of bad API; s_umount is used to pin
196the superblock down when we only have dev_t given us by userland to
197identify the superblock. Everything else (statfs(), fstatfs(), etc.)
198doesn't hold it when calling ->statfs() - superblock is pinned down
199by resolving the pathname passed to syscall.
200
201->quota_read() and ->quota_write() functions are both guaranteed to
202be the only ones operating on the quota file by the quota code (via
203dqio_sem) (unless an admin really wants to screw up something and
204writes to quota files with quotas on). For other details about locking
205see also dquot_operations section.
206
207->bdev_try_to_free_page is called from the ->releasepage handler of
208the block device inode. See there for more details.
209
210file_system_type
211================
212
213prototypes::
214
215 struct dentry *(*mount) (struct file_system_type *, int,
216 const char *, void *);
217 void (*kill_sb) (struct super_block *);
218
219locking rules:
220
221======= =========
222ops may block
223======= =========
224mount yes
225kill_sb yes
226======= =========
227
228->mount() returns ERR_PTR or the root dentry; its superblock should be locked
229on return.
230
231->kill_sb() takes a write-locked superblock, does all shutdown work on it,
232unlocks and drops the reference.
233
234address_space_operations
235========================
236prototypes::
237
238 int (*writepage)(struct page *page, struct writeback_control *wbc);
239 int (*readpage)(struct file *, struct page *);
240 int (*writepages)(struct address_space *, struct writeback_control *);
241 int (*set_page_dirty)(struct page *page);
242 int (*readpages)(struct file *filp, struct address_space *mapping,
243 struct list_head *pages, unsigned nr_pages);
244 int (*write_begin)(struct file *, struct address_space *mapping,
245 loff_t pos, unsigned len, unsigned flags,
246 struct page **pagep, void **fsdata);
247 int (*write_end)(struct file *, struct address_space *mapping,
248 loff_t pos, unsigned len, unsigned copied,
249 struct page *page, void *fsdata);
250 sector_t (*bmap)(struct address_space *, sector_t);
251 void (*invalidatepage) (struct page *, unsigned int, unsigned int);
252 int (*releasepage) (struct page *, int);
253 void (*freepage)(struct page *);
254 int (*direct_IO)(struct kiocb *, struct iov_iter *iter);
255 bool (*isolate_page) (struct page *, isolate_mode_t);
256 int (*migratepage)(struct address_space *, struct page *, struct page *);
257 void (*putback_page) (struct page *);
258 int (*launder_page)(struct page *);
259 int (*is_partially_uptodate)(struct page *, unsigned long, unsigned long);
260 int (*error_remove_page)(struct address_space *, struct page *);
261 int (*swap_activate)(struct file *);
262 int (*swap_deactivate)(struct file *);
263
264locking rules:
265 All except set_page_dirty and freepage may block
266
267====================== ======================== =========
268ops PageLocked(page) i_rwsem
269====================== ======================== =========
270writepage: yes, unlocks (see below)
271readpage: yes, unlocks
272writepages:
273set_page_dirty no
274readpages:
275write_begin: locks the page exclusive
276write_end: yes, unlocks exclusive
277bmap:
278invalidatepage: yes
279releasepage: yes
280freepage: yes
281direct_IO:
282isolate_page: yes
283migratepage: yes (both)
284putback_page: yes
285launder_page: yes
286is_partially_uptodate: yes
287error_remove_page: yes
288swap_activate: no
289swap_deactivate: no
290====================== ======================== =========
291
292->write_begin(), ->write_end() and ->readpage() may be called from
293the request handler (/dev/loop).
294
295->readpage() unlocks the page, either synchronously or via I/O
296completion.
297
298->readpages() populates the pagecache with the passed pages and starts
299I/O against them. They come unlocked upon I/O completion.
300
301->writepage() is used for two purposes: for "memory cleansing" and for
302"sync". These are quite different operations and the behaviour may differ
303depending upon the mode.
304
305If writepage is called for sync (wbc->sync_mode != WBC_SYNC_NONE) then
306it *must* start I/O against the page, even if that would involve
307blocking on in-progress I/O.
308
309If writepage is called for memory cleansing (sync_mode ==
310WBC_SYNC_NONE) then its role is to get as much writeout underway as
311possible. So writepage should try to avoid blocking against
312currently-in-progress I/O.
313
314If the filesystem is not called for "sync" and it determines that it
315would need to block against in-progress I/O to be able to start new I/O
316against the page the filesystem should redirty the page with
317redirty_page_for_writepage(), then unlock the page and return zero.
318This may also be done to avoid internal deadlocks, but rarely.
319
320If the filesystem is called for sync then it must wait on any
321in-progress I/O and then start new I/O.
322
323The filesystem should unlock the page synchronously, before returning to the
324caller, unless ->writepage() returns special WRITEPAGE_ACTIVATE
325value. WRITEPAGE_ACTIVATE means that page cannot really be written out
326currently, and VM should stop calling ->writepage() on this page for some
327time. VM does this by moving page to the head of the active list, hence the
328name.
329
330Unless the filesystem is going to redirty_page_for_writepage(), unlock the page
331and return zero, writepage *must* run set_page_writeback() against the page,
332followed by unlocking it. Once set_page_writeback() has been run against the
333page, write I/O can be submitted and the write I/O completion handler must run
334end_page_writeback() once the I/O is complete. If no I/O is submitted, the
335filesystem must run end_page_writeback() against the page before returning from
336writepage.
337
338That is: after 2.5.12, pages which are under writeout are *not* locked. Note,
339if the filesystem needs the page to be locked during writeout, that is ok, too,
340the page is allowed to be unlocked at any point in time between the calls to
341set_page_writeback() and end_page_writeback().
342
343Note, failure to run either redirty_page_for_writepage() or the combination of
344set_page_writeback()/end_page_writeback() on a page submitted to writepage
345will leave the page itself marked clean but it will be tagged as dirty in the
346radix tree. This incoherency can lead to all sorts of hard-to-debug problems
347in the filesystem like having dirty inodes at umount and losing written data.
348
349->writepages() is used for periodic writeback and for syscall-initiated
350sync operations. The address_space should start I/O against at least
351``*nr_to_write`` pages. ``*nr_to_write`` must be decremented for each page
352which is written. The address_space implementation may write more (or less)
353pages than ``*nr_to_write`` asks for, but it should try to be reasonably close.
354If nr_to_write is NULL, all dirty pages must be written.
355
356writepages should _only_ write pages which are present on
357mapping->io_pages.
358
359->set_page_dirty() is called from various places in the kernel
360when the target page is marked as needing writeback. It may be called
361under spinlock (it cannot block) and is sometimes called with the page
362not locked.
363
364->bmap() is currently used by legacy ioctl() (FIBMAP) provided by some
365filesystems and by the swapper. The latter will eventually go away. Please,
366keep it that way and don't breed new callers.
367
368->invalidatepage() is called when the filesystem must attempt to drop
369some or all of the buffers from the page when it is being truncated. It
370returns zero on success. If ->invalidatepage is zero, the kernel uses
371block_invalidatepage() instead.
372
373->releasepage() is called when the kernel is about to try to drop the
374buffers from the page in preparation for freeing it. It returns zero to
375indicate that the buffers are (or may be) freeable. If ->releasepage is zero,
376the kernel assumes that the fs has no private interest in the buffers.
377
378->freepage() is called when the kernel is done dropping the page
379from the page cache.
380
381->launder_page() may be called prior to releasing a page if
382it is still found to be dirty. It returns zero if the page was successfully
383cleaned, or an error value if not. Note that in order to prevent the page
384getting mapped back in and redirtied, it needs to be kept locked
385across the entire operation.
386
387->swap_activate will be called with a non-zero argument on
388files backing (non block device backed) swapfiles. A return value
389of zero indicates success, in which case this file can be used for
390backing swapspace. The swapspace operations will be proxied to the
391address space operations.
392
393->swap_deactivate() will be called in the sys_swapoff()
394path after ->swap_activate() returned success.
395
396file_lock_operations
397====================
398
399prototypes::
400
401 void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
402 void (*fl_release_private)(struct file_lock *);
403
404
405locking rules:
406
407=================== ============= =========
408ops inode->i_lock may block
409=================== ============= =========
410fl_copy_lock: yes no
411fl_release_private: maybe maybe[1]_
412=================== ============= =========
413
414.. [1]:
415 ->fl_release_private for flock or POSIX locks is currently allowed
416 to block. Leases however can still be freed while the i_lock is held and
417 so fl_release_private called on a lease should not block.
418
419lock_manager_operations
420=======================
421
422prototypes::
423
424 void (*lm_notify)(struct file_lock *); /* unblock callback */
425 int (*lm_grant)(struct file_lock *, struct file_lock *, int);
426 void (*lm_break)(struct file_lock *); /* break_lease callback */
427 int (*lm_change)(struct file_lock **, int);
428
429locking rules:
430
431========== ============= ================= =========
432ops inode->i_lock blocked_lock_lock may block
433========== ============= ================= =========
434lm_notify: yes yes no
435lm_grant: no no no
436lm_break: yes no no
437lm_change yes no no
438========== ============= ================= =========
439
440buffer_head
441===========
442
443prototypes::
444
445 void (*b_end_io)(struct buffer_head *bh, int uptodate);
446
447locking rules:
448
449called from interrupts. In other words, extreme care is needed here.
450bh is locked, but that's all warranties we have here. Currently only RAID1,
451highmem, fs/buffer.c, and fs/ntfs/aops.c are providing these. Block devices
452call this method upon the IO completion.
453
454block_device_operations
455=======================
456prototypes::
457
458 int (*open) (struct block_device *, fmode_t);
459 int (*release) (struct gendisk *, fmode_t);
460 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
461 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
462 int (*direct_access) (struct block_device *, sector_t, void **,
463 unsigned long *);
464 int (*media_changed) (struct gendisk *);
465 void (*unlock_native_capacity) (struct gendisk *);
466 int (*revalidate_disk) (struct gendisk *);
467 int (*getgeo)(struct block_device *, struct hd_geometry *);
468 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
469
470locking rules:
471
472======================= ===================
473ops bd_mutex
474======================= ===================
475open: yes
476release: yes
477ioctl: no
478compat_ioctl: no
479direct_access: no
480media_changed: no
481unlock_native_capacity: no
482revalidate_disk: no
483getgeo: no
484swap_slot_free_notify: no (see below)
485======================= ===================
486
487media_changed, unlock_native_capacity and revalidate_disk are called only from
488check_disk_change().
489
490swap_slot_free_notify is called with swap_lock and sometimes the page lock
491held.
492
493
494file_operations
495===============
496
497prototypes::
498
499 loff_t (*llseek) (struct file *, loff_t, int);
500 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
501 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
502 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
503 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
504 int (*iterate) (struct file *, struct dir_context *);
505 int (*iterate_shared) (struct file *, struct dir_context *);
506 __poll_t (*poll) (struct file *, struct poll_table_struct *);
507 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
508 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
509 int (*mmap) (struct file *, struct vm_area_struct *);
510 int (*open) (struct inode *, struct file *);
511 int (*flush) (struct file *);
512 int (*release) (struct inode *, struct file *);
513 int (*fsync) (struct file *, loff_t start, loff_t end, int datasync);
514 int (*fasync) (int, struct file *, int);
515 int (*lock) (struct file *, int, struct file_lock *);
516 ssize_t (*readv) (struct file *, const struct iovec *, unsigned long,
517 loff_t *);
518 ssize_t (*writev) (struct file *, const struct iovec *, unsigned long,
519 loff_t *);
520 ssize_t (*sendfile) (struct file *, loff_t *, size_t, read_actor_t,
521 void __user *);
522 ssize_t (*sendpage) (struct file *, struct page *, int, size_t,
523 loff_t *, int);
524 unsigned long (*get_unmapped_area)(struct file *, unsigned long,
525 unsigned long, unsigned long, unsigned long);
526 int (*check_flags)(int);
527 int (*flock) (struct file *, int, struct file_lock *);
528 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *,
529 size_t, unsigned int);
530 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *,
531 size_t, unsigned int);
532 int (*setlease)(struct file *, long, struct file_lock **, void **);
533 long (*fallocate)(struct file *, int, loff_t, loff_t);
534
535locking rules:
536 All may block.
537
538->llseek() locking has moved from llseek to the individual llseek
539implementations. If your fs is not using generic_file_llseek, you
540need to acquire and release the appropriate locks in your ->llseek().
541For many filesystems, it is probably safe to acquire the inode
542mutex or just to use i_size_read() instead.
543Note: this does not protect the file->f_pos against concurrent modifications
544since this is something the userspace has to take care about.
545
546->iterate() is called with i_rwsem exclusive.
547
548->iterate_shared() is called with i_rwsem at least shared.
549
550->fasync() is responsible for maintaining the FASYNC bit in filp->f_flags.
551Most instances call fasync_helper(), which does that maintenance, so it's
552not normally something one needs to worry about. Return values > 0 will be
553mapped to zero in the VFS layer.
554
555->readdir() and ->ioctl() on directories must be changed. Ideally we would
556move ->readdir() to inode_operations and use a separate method for directory
557->ioctl() or kill the latter completely. One of the problems is that for
558anything that resembles union-mount we won't have a struct file for all
559components. And there are other reasons why the current interface is a mess...
560
561->read on directories probably must go away - we should just enforce -EISDIR
562in sys_read() and friends.
563
564->setlease operations should call generic_setlease() before or after setting
565the lease within the individual filesystem to record the result of the
566operation
567
568dquot_operations
569================
570
571prototypes::
572
573 int (*write_dquot) (struct dquot *);
574 int (*acquire_dquot) (struct dquot *);
575 int (*release_dquot) (struct dquot *);
576 int (*mark_dirty) (struct dquot *);
577 int (*write_info) (struct super_block *, int);
578
579These operations are intended to be more or less wrapping functions that ensure
580a proper locking wrt the filesystem and call the generic quota operations.
581
582What filesystem should expect from the generic quota functions:
583
584============== ============ =========================
585ops FS recursion Held locks when called
586============== ============ =========================
587write_dquot: yes dqonoff_sem or dqptr_sem
588acquire_dquot: yes dqonoff_sem or dqptr_sem
589release_dquot: yes dqonoff_sem or dqptr_sem
590mark_dirty: no -
591write_info: yes dqonoff_sem
592============== ============ =========================
593
594FS recursion means calling ->quota_read() and ->quota_write() from superblock
595operations.
596
597More details about quota locking can be found in fs/dquot.c.
598
599vm_operations_struct
600====================
601
602prototypes::
603
604 void (*open)(struct vm_area_struct*);
605 void (*close)(struct vm_area_struct*);
606 vm_fault_t (*fault)(struct vm_area_struct*, struct vm_fault *);
607 vm_fault_t (*page_mkwrite)(struct vm_area_struct *, struct vm_fault *);
608 vm_fault_t (*pfn_mkwrite)(struct vm_area_struct *, struct vm_fault *);
609 int (*access)(struct vm_area_struct *, unsigned long, void*, int, int);
610
611locking rules:
612
613============= ======== ===========================
614ops mmap_sem PageLocked(page)
615============= ======== ===========================
616open: yes
617close: yes
618fault: yes can return with page locked
619map_pages: yes
620page_mkwrite: yes can return with page locked
621pfn_mkwrite: yes
622access: yes
623============= ======== ===========================
624
625->fault() is called when a previously not present pte is about
626to be faulted in. The filesystem must find and return the page associated
627with the passed in "pgoff" in the vm_fault structure. If it is possible that
628the page may be truncated and/or invalidated, then the filesystem must lock
629the page, then ensure it is not already truncated (the page lock will block
630subsequent truncate), and then return with VM_FAULT_LOCKED, and the page
631locked. The VM will unlock the page.
632
633->map_pages() is called when VM asks to map easy accessible pages.
634Filesystem should find and map pages associated with offsets from "start_pgoff"
635till "end_pgoff". ->map_pages() is called with page table locked and must
636not block. If it's not possible to reach a page without blocking,
637filesystem should skip it. Filesystem should use do_set_pte() to setup
638page table entry. Pointer to entry associated with the page is passed in
639"pte" field in vm_fault structure. Pointers to entries for other offsets
640should be calculated relative to "pte".
641
642->page_mkwrite() is called when a previously read-only pte is
643about to become writeable. The filesystem again must ensure that there are
644no truncate/invalidate races, and then return with the page locked. If
645the page has been truncated, the filesystem should not look up a new page
646like the ->fault() handler, but simply return with VM_FAULT_NOPAGE, which
647will cause the VM to retry the fault.
648
649->pfn_mkwrite() is the same as page_mkwrite but when the pte is
650VM_PFNMAP or VM_MIXEDMAP with a page-less entry. Expected return is
651VM_FAULT_NOPAGE. Or one of the VM_FAULT_ERROR types. The default behavior
652after this call is to make the pte read-write, unless pfn_mkwrite returns
653an error.
654
655->access() is called when get_user_pages() fails in
656access_process_vm(), typically used to debug a process through
657/proc/pid/mem or ptrace. This function is needed only for
658VM_IO | VM_PFNMAP VMAs.
659
660--------------------------------------------------------------------------------
661
662 Dubious stuff
663
664(if you break something or notice that it is broken and do not fix it yourself
665- at least put it here)