Documentation/filesystems/Locking

   1         The text below describes the locking rules for VFS-related methods.
   2 It is (believed to be) up-to-date. *Please*, if you change anything in
   3 prototypes or locking protocols - update this file. And update the relevant
   4 instances in the tree, don't leave that to maintainers of filesystems/devices/
   5 etc. At the very least, put the list of dubious cases in the end of this file.
   6 Don't turn it into log - maintainers of out-of-the-tree code are supposed to
   7 be able to use diff(1).
   8         Thing currently missing here: socket operations. Alexey?
   9
  10 --------------------------- dentry_operations --------------------------
  11 prototypes:
  12         int (*d_revalidate)(struct dentry *, struct nameidata *);
  13         int (*d_hash)(const struct dentry *, const struct inode *,
  14                         struct qstr *);
  15         int (*d_compare)(const struct dentry *, const struct inode *,
  16                         const struct dentry *, const struct inode *,
  17                         unsigned int, const char *, const struct qstr *);
  18         int (*d_delete)(struct dentry *);
  19         void (*d_release)(struct dentry *);
  20         void (*d_iput)(struct dentry *, struct inode *);
  21         char *(*d_dname)((struct dentry *dentry, char *buffer, int buflen);
  22
  23 locking rules:
  24                 rename_lock     ->d_lock        may block       rcu-walk
  25 d_revalidate:   no              no              yes (ref-walk)  maybe
  26 d_hash          no              no              no              maybe
  27 d_compare:      yes             no              no              maybe
  28 d_delete:       no              yes             no              no
  29 d_release:      no              no              yes             no
  30 d_iput:         no              no              yes             no
  31 d_dname:        no              no              no              no
  32
  33 --------------------------- inode_operations ---------------------------
  34 prototypes:
  35         int (*create) (struct inode *,struct dentry *,int, struct nameidata *);
  36         struct dentry * (*lookup) (struct inode *,struct dentry *, struct nameid
  37 ata *);
  38         int (*link) (struct dentry *,struct inode *,struct dentry *);
  39         int (*unlink) (struct inode *,struct dentry *);
  40         int (*symlink) (struct inode *,struct dentry *,const char *);
  41         int (*mkdir) (struct inode *,struct dentry *,int);
  42         int (*rmdir) (struct inode *,struct dentry *);
  43         int (*mknod) (struct inode *,struct dentry *,int,dev_t);
  44         int (*rename) (struct inode *, struct dentry *,
  45                         struct inode *, struct dentry *);
  46         int (*readlink) (struct dentry *, char __user *,int);
  47         void * (*follow_link) (struct dentry *, struct nameidata *);
  48         void (*put_link) (struct dentry *, struct nameidata *, void *);
  49         void (*truncate) (struct inode *);
  50         int (*permission) (struct inode *, int, unsigned int);
  51         int (*check_acl)(struct inode *, int, unsigned int);
  52         int (*setattr) (struct dentry *, struct iattr *);
  53         int (*getattr) (struct vfsmount *, struct dentry *, struct kstat *);
  54         int (*setxattr) (struct dentry *, const char *,const void *,size_t,int);
  55         ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t);
  56         ssize_t (*listxattr) (struct dentry *, char *, size_t);
  57         int (*removexattr) (struct dentry *, const char *);
  58         void (*truncate_range)(struct inode *, loff_t, loff_t);
  59         long (*fallocate)(struct inode *inode, int mode, loff_t offset, loff_t len);
  60         int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start, u64 len);
  61
  62 locking rules:
  63         all may block
  64                 i_mutex(inode)
  65 lookup:         yes
  66 create:         yes
  67 link:           yes (both)
  68 mknod:          yes
  69 symlink:        yes
  70 mkdir:          yes
  71 unlink:         yes (both)
  72 rmdir:          yes (both)      (see below)
  73 rename:         yes (all)       (see below)
  74 readlink:       no
  75 follow_link:    no
  76 put_link:       no
  77 truncate:       yes             (see below)
  78 setattr:        yes
  79 permission:     no (may not block if called in rcu-walk mode)
  80 check_acl:      no
  81 getattr:        no
  82 setxattr:       yes
  83 getxattr:       no
  84 listxattr:      no
  85 removexattr:    yes
  86 truncate_range: yes
  87 fallocate:      no
  88 fiemap:         no
  89         Additionally, ->rmdir(), ->unlink() and ->rename() have ->i_mutex on
  90 victim.
  91         cross-directory ->rename() has (per-superblock) ->s_vfs_rename_sem.
  92         ->truncate() is never called directly - it's a callback, not a
  93 method. It's called by vmtruncate() - deprecated library function used by
  94 ->setattr(). Locking information above applies to that call (i.e. is
  95 inherited from ->setattr() - vmtruncate() is used when ATTR_SIZE had been
  96 passed).
  97
  98 See Documentation/filesystems/directory-locking for more detailed discussion
  99 of the locking scheme for directory operations.
 100
 101 --------------------------- super_operations ---------------------------
 102 prototypes:
 103         struct inode *(*alloc_inode)(struct super_block *sb);
 104         void (*destroy_inode)(struct inode *);
 105         void (*dirty_inode) (struct inode *);
 106         int (*write_inode) (struct inode *, struct writeback_control *wbc);
 107         int (*drop_inode) (struct inode *);
 108         void (*evict_inode) (struct inode *);
 109         void (*put_super) (struct super_block *);
 110         void (*write_super) (struct super_block *);
 111         int (*sync_fs)(struct super_block *sb, int wait);
 112         int (*freeze_fs) (struct super_block *);
 113         int (*unfreeze_fs) (struct super_block *);
 114         int (*statfs) (struct dentry *, struct kstatfs *);
 115         int (*remount_fs) (struct super_block *, int *, char *);
 116         void (*umount_begin) (struct super_block *);
 117         int (*show_options)(struct seq_file *, struct vfsmount *);
 118         ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
 119         ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
 120         int (*bdev_try_to_free_page)(struct super_block*, struct page*, gfp_t);
 121
 122 locking rules:
 123         All may block [not true, see below]
 124                         s_umount
 125 alloc_inode:
 126 destroy_inode:
 127 dirty_inode:                            (must not sleep)
 128 write_inode:
 129 drop_inode:                             !!!inode_lock!!!
 130 evict_inode:
 131 put_super:              write
 132 write_super:            read
 133 sync_fs:                read
 134 freeze_fs:              read
 135 unfreeze_fs:            read
 136 statfs:                 maybe(read)     (see below)
 137 remount_fs:             write
 138 umount_begin:           no
 139 show_options:           no              (namespace_sem)
 140 quota_read:             no              (see below)
 141 quota_write:            no              (see below)
 142 bdev_try_to_free_page:  no              (see below)
 143
 144 ->statfs() has s_umount (shared) when called by ustat(2) (native or
 145 compat), but that's an accident of bad API; s_umount is used to pin
 146 the superblock down when we only have dev_t given us by userland to
 147 identify the superblock.  Everything else (statfs(), fstatfs(), etc.)
 148 doesn't hold it when calling ->statfs() - superblock is pinned down
 149 by resolving the pathname passed to syscall.
 150 ->quota_read() and ->quota_write() functions are both guaranteed to
 151 be the only ones operating on the quota file by the quota code (via
 152 dqio_sem) (unless an admin really wants to screw up something and
 153 writes to quota files with quotas on). For other details about locking
 154 see also dquot_operations section.
 155 ->bdev_try_to_free_page is called from the ->releasepage handler of
 156 the block device inode.  See there for more details.
 157
 158 --------------------------- file_system_type ---------------------------
 159 prototypes:
 160         int (*get_sb) (struct file_system_type *, int,
 161                        const char *, void *, struct vfsmount *);
 162         struct dentry *(*mount) (struct file_system_type *, int,
 163                        const char *, void *);
 164         void (*kill_sb) (struct super_block *);
 165 locking rules:
 166                 may block
 167 get_sb          yes
 168 mount           yes
 169 kill_sb         yes
 170
 171 ->get_sb() returns error or 0 with locked superblock attached to the vfsmount
 172 (exclusive on ->s_umount).
 173 ->mount() returns ERR_PTR or the root dentry.
 174 ->kill_sb() takes a write-locked superblock, does all shutdown work on it,
 175 unlocks and drops the reference.
 176
 177 --------------------------- address_space_operations --------------------------
 178 prototypes:
 179         int (*writepage)(struct page *page, struct writeback_control *wbc);
 180         int (*readpage)(struct file *, struct page *);
 181         int (*sync_page)(struct page *);
 182         int (*writepages)(struct address_space *, struct writeback_control *);
 183         int (*set_page_dirty)(struct page *page);
 184         int (*readpages)(struct file *filp, struct address_space *mapping,
 185                         struct list_head *pages, unsigned nr_pages);
 186         int (*write_begin)(struct file *, struct address_space *mapping,
 187                                 loff_t pos, unsigned len, unsigned flags,
 188                                 struct page **pagep, void **fsdata);
 189         int (*write_end)(struct file *, struct address_space *mapping,
 190                                 loff_t pos, unsigned len, unsigned copied,
 191                                 struct page *page, void *fsdata);
 192         sector_t (*bmap)(struct address_space *, sector_t);
 193         int (*invalidatepage) (struct page *, unsigned long);
 194         int (*releasepage) (struct page *, int);
 195         void (*freepage)(struct page *);
 196         int (*direct_IO)(int, struct kiocb *, const struct iovec *iov,
 197                         loff_t offset, unsigned long nr_segs);
 198         int (*get_xip_mem)(struct address_space *, pgoff_t, int, void **,
 199                                 unsigned long *);
 200         int (*migratepage)(struct address_space *, struct page *, struct page *);
 201         int (*launder_page)(struct page *);
 202         int (*is_partially_uptodate)(struct page *, read_descriptor_t *, unsigned long);
 203         int (*error_remove_page)(struct address_space *, struct page *);
 204
 205 locking rules:
 206         All except set_page_dirty and freepage may block
 207
 208                         PageLocked(page)        i_mutex
 209 writepage:              yes, unlocks (see below)
 210 readpage:               yes, unlocks
 211 sync_page:              maybe
 212 writepages:
 213 set_page_dirty          no
 214 readpages:
 215 write_begin:            locks the page          yes
 216 write_end:              yes, unlocks            yes
 217 bmap:
 218 invalidatepage:         yes
 219 releasepage:            yes
 220 freepage:               yes
 221 direct_IO:
 222 get_xip_mem:                                    maybe
 223 migratepage:            yes (both)
 224 launder_page:           yes
 225 is_partially_uptodate:  yes
 226 error_remove_page:      yes
 227
 228         ->write_begin(), ->write_end(), ->sync_page() and ->readpage()
 229 may be called from the request handler (/dev/loop).
 230
 231         ->readpage() unlocks the page, either synchronously or via I/O
 232 completion.
 233
 234         ->readpages() populates the pagecache with the passed pages and starts
 235 I/O against them.  They come unlocked upon I/O completion.
 236
 237         ->writepage() is used for two purposes: for "memory cleansing" and for
 238 "sync".  These are quite different operations and the behaviour may differ
 239 depending upon the mode.
 240
 241 If writepage is called for sync (wbc->sync_mode != WBC_SYNC_NONE) then
 242 it *must* start I/O against the page, even if that would involve
 243 blocking on in-progress I/O.
 244
 245 If writepage is called for memory cleansing (sync_mode ==
 246 WBC_SYNC_NONE) then its role is to get as much writeout underway as
 247 possible.  So writepage should try to avoid blocking against
 248 currently-in-progress I/O.
 249
 250 If the filesystem is not called for "sync" and it determines that it
 251 would need to block against in-progress I/O to be able to start new I/O
 252 against the page the filesystem should redirty the page with
 253 redirty_page_for_writepage(), then unlock the page and return zero.
 254 This may also be done to avoid internal deadlocks, but rarely.
 255
 256 If the filesystem is called for sync then it must wait on any
 257 in-progress I/O and then start new I/O.
 258
 259 The filesystem should unlock the page synchronously, before returning to the
 260 caller, unless ->writepage() returns special WRITEPAGE_ACTIVATE
 261 value. WRITEPAGE_ACTIVATE means that page cannot really be written out
 262 currently, and VM should stop calling ->writepage() on this page for some
 263 time. VM does this by moving page to the head of the active list, hence the
 264 name.
 265
 266 Unless the filesystem is going to redirty_page_for_writepage(), unlock the page
 267 and return zero, writepage *must* run set_page_writeback() against the page,
 268 followed by unlocking it.  Once set_page_writeback() has been run against the
 269 page, write I/O can be submitted and the write I/O completion handler must run
 270 end_page_writeback() once the I/O is complete.  If no I/O is submitted, the
 271 filesystem must run end_page_writeback() against the page before returning from
 272 writepage.
 273
 274 That is: after 2.5.12, pages which are under writeout are *not* locked.  Note,
 275 if the filesystem needs the page to be locked during writeout, that is ok, too,
 276 the page is allowed to be unlocked at any point in time between the calls to
 277 set_page_writeback() and end_page_writeback().
 278
 279 Note, failure to run either redirty_page_for_writepage() or the combination of
 280 set_page_writeback()/end_page_writeback() on a page submitted to writepage
 281 will leave the page itself marked clean but it will be tagged as dirty in the
 282 radix tree.  This incoherency can lead to all sorts of hard-to-debug problems
 283 in the filesystem like having dirty inodes at umount and losing written data.
 284
 285         ->sync_page() locking rules are not well-defined - usually it is called
 286 with lock on page, but that is not guaranteed. Considering the currently
 287 existing instances of this method ->sync_page() itself doesn't look
 288 well-defined...
 289
 290         ->writepages() is used for periodic writeback and for syscall-initiated
 291 sync operations.  The address_space should start I/O against at least
 292 *nr_to_write pages.  *nr_to_write must be decremented for each page which is
 293 written.  The address_space implementation may write more (or less) pages
 294 than *nr_to_write asks for, but it should try to be reasonably close.  If
 295 nr_to_write is NULL, all dirty pages must be written.
 296
 297 writepages should _only_ write pages which are present on
 298 mapping->io_pages.
 299
 300         ->set_page_dirty() is called from various places in the kernel
 301 when the target page is marked as needing writeback.  It may be called
 302 under spinlock (it cannot block) and is sometimes called with the page
 303 not locked.
 304
 305         ->bmap() is currently used by legacy ioctl() (FIBMAP) provided by some
 306 filesystems and by the swapper. The latter will eventually go away.  Please,
 307 keep it that way and don't breed new callers.
 308
 309         ->invalidatepage() is called when the filesystem must attempt to drop
 310 some or all of the buffers from the page when it is being truncated.  It
 311 returns zero on success.  If ->invalidatepage is zero, the kernel uses
 312 block_invalidatepage() instead.
 313
 314         ->releasepage() is called when the kernel is about to try to drop the
 315 buffers from the page in preparation for freeing it.  It returns zero to
 316 indicate that the buffers are (or may be) freeable.  If ->releasepage is zero,
 317 the kernel assumes that the fs has no private interest in the buffers.
 318
 319         ->freepage() is called when the kernel is done dropping the page
 320 from the page cache.
 321
 322         ->launder_page() may be called prior to releasing a page if
 323 it is still found to be dirty. It returns zero if the page was successfully
 324 cleaned, or an error value if not. Note that in order to prevent the page
 325 getting mapped back in and redirtied, it needs to be kept locked
 326 across the entire operation.
 327
 328 ----------------------- file_lock_operations ------------------------------
 329 prototypes:
 330         void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
 331         void (*fl_release_private)(struct file_lock *);
 332
 333
 334 locking rules:
 335                         file_lock_lock  may block
 336 fl_copy_lock:           yes             no
 337 fl_release_private:     maybe           no
 338
 339 ----------------------- lock_manager_operations ---------------------------
 340 prototypes:
 341         int (*fl_compare_owner)(struct file_lock *, struct file_lock *);
 342         void (*fl_notify)(struct file_lock *);  /* unblock callback */
 343         int (*fl_grant)(struct file_lock *, struct file_lock *, int);
 344         void (*fl_release_private)(struct file_lock *);
 345         void (*fl_break)(struct file_lock *); /* break_lease callback */
 346         int (*fl_mylease)(struct file_lock *, struct file_lock *);
 347         int (*fl_change)(struct file_lock **, int);
 348
 349 locking rules:
 350                         file_lock_lock  may block
 351 fl_compare_owner:       yes             no
 352 fl_notify:              yes             no
 353 fl_grant:               no              no
 354 fl_release_private:     maybe           no
 355 fl_break:               yes             no
 356 fl_mylease:             yes             no
 357 fl_change               yes             no
 358
 359 --------------------------- buffer_head -----------------------------------
 360 prototypes:
 361         void (*b_end_io)(struct buffer_head *bh, int uptodate);
 362
 363 locking rules:
 364         called from interrupts. In other words, extreme care is needed here.
 365 bh is locked, but that's all warranties we have here. Currently only RAID1,
 366 highmem, fs/buffer.c, and fs/ntfs/aops.c are providing these. Block devices
 367 call this method upon the IO completion.
 368
 369 --------------------------- block_device_operations -----------------------
 370 prototypes:
 371         int (*open) (struct block_device *, fmode_t);
 372         int (*release) (struct gendisk *, fmode_t);
 373         int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
 374         int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long);
 375         int (*direct_access) (struct block_device *, sector_t, void **, unsigned long *);
 376         int (*media_changed) (struct gendisk *);
 377         void (*unlock_native_capacity) (struct gendisk *);
 378         int (*revalidate_disk) (struct gendisk *);
 379         int (*getgeo)(struct block_device *, struct hd_geometry *);
 380         void (*swap_slot_free_notify) (struct block_device *, unsigned long);
 381
 382 locking rules:
 383                         bd_mutex
 384 open:                   yes
 385 release:                yes
 386 ioctl:                  no
 387 compat_ioctl:           no
 388 direct_access:          no
 389 media_changed:          no
 390 unlock_native_capacity: no
 391 revalidate_disk:        no
 392 getgeo:                 no
 393 swap_slot_free_notify:  no      (see below)
 394
 395 media_changed, unlock_native_capacity and revalidate_disk are called only from
 396 check_disk_change().
 397
 398 swap_slot_free_notify is called with swap_lock and sometimes the page lock
 399 held.
 400
 401
 402 --------------------------- file_operations -------------------------------
 403 prototypes:
 404         loff_t (*llseek) (struct file *, loff_t, int);
 405         ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
 406         ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
 407         ssize_t (*aio_read) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
 408         ssize_t (*aio_write) (struct kiocb *, const struct iovec *, unsigned long, loff_t);
 409         int (*readdir) (struct file *, void *, filldir_t);
 410         unsigned int (*poll) (struct file *, struct poll_table_struct *);
 411         long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
 412         long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
 413         int (*mmap) (struct file *, struct vm_area_struct *);
 414         int (*open) (struct inode *, struct file *);
 415         int (*flush) (struct file *);
 416         int (*release) (struct inode *, struct file *);
 417         int (*fsync) (struct file *, int datasync);
 418         int (*aio_fsync) (struct kiocb *, int datasync);
 419         int (*fasync) (int, struct file *, int);
 420         int (*lock) (struct file *, int, struct file_lock *);
 421         ssize_t (*readv) (struct file *, const struct iovec *, unsigned long,
 422                         loff_t *);
 423         ssize_t (*writev) (struct file *, const struct iovec *, unsigned long,
 424                         loff_t *);
 425         ssize_t (*sendfile) (struct file *, loff_t *, size_t, read_actor_t,
 426                         void __user *);
 427         ssize_t (*sendpage) (struct file *, struct page *, int, size_t,
 428                         loff_t *, int);
 429         unsigned long (*get_unmapped_area)(struct file *, unsigned long,
 430                         unsigned long, unsigned long, unsigned long);
 431         int (*check_flags)(int);
 432         int (*flock) (struct file *, int, struct file_lock *);
 433         ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *,
 434                         size_t, unsigned int);
 435         ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *,
 436                         size_t, unsigned int);
 437         int (*setlease)(struct file *, long, struct file_lock **);
 438 };
 439
 440 locking rules:
 441         All may block except for ->setlease.
 442         No VFS locks held on entry except for ->fsync and ->setlease.
 443
 444 ->fsync() has i_mutex on inode.
 445
 446 ->setlease has the file_list_lock held and must not sleep.
 447
 448 ->llseek() locking has moved from llseek to the individual llseek
 449 implementations.  If your fs is not using generic_file_llseek, you
 450 need to acquire and release the appropriate locks in your ->llseek().
 451 For many filesystems, it is probably safe to acquire the inode
 452 mutex or just to use i_size_read() instead.
 453 Note: this does not protect the file->f_pos against concurrent modifications
 454 since this is something the userspace has to take care about.
 455
 456 ->fasync() is responsible for maintaining the FASYNC bit in filp->f_flags.
 457 Most instances call fasync_helper(), which does that maintenance, so it's
 458 not normally something one needs to worry about.  Return values > 0 will be
 459 mapped to zero in the VFS layer.
 460
 461 ->readdir() and ->ioctl() on directories must be changed. Ideally we would
 462 move ->readdir() to inode_operations and use a separate method for directory
 463 ->ioctl() or kill the latter completely. One of the problems is that for
 464 anything that resembles union-mount we won't have a struct file for all
 465 components. And there are other reasons why the current interface is a mess...
 466
 467 ->read on directories probably must go away - we should just enforce -EISDIR
 468 in sys_read() and friends.
 469
 470 --------------------------- dquot_operations -------------------------------
 471 prototypes:
 472         int (*write_dquot) (struct dquot *);
 473         int (*acquire_dquot) (struct dquot *);
 474         int (*release_dquot) (struct dquot *);
 475         int (*mark_dirty) (struct dquot *);
 476         int (*write_info) (struct super_block *, int);
 477
 478 These operations are intended to be more or less wrapping functions that ensure
 479 a proper locking wrt the filesystem and call the generic quota operations.
 480
 481 What filesystem should expect from the generic quota functions:
 482
 483                 FS recursion    Held locks when called
 484 write_dquot:    yes             dqonoff_sem or dqptr_sem
 485 acquire_dquot:  yes             dqonoff_sem or dqptr_sem
 486 release_dquot:  yes             dqonoff_sem or dqptr_sem
 487 mark_dirty:     no              -
 488 write_info:     yes             dqonoff_sem
 489
 490 FS recursion means calling ->quota_read() and ->quota_write() from superblock
 491 operations.
 492
 493 More details about quota locking can be found in fs/dquot.c.
 494
 495 --------------------------- vm_operations_struct -----------------------------
 496 prototypes:
 497         void (*open)(struct vm_area_struct*);
 498         void (*close)(struct vm_area_struct*);
 499         int (*fault)(struct vm_area_struct*, struct vm_fault *);
 500         int (*page_mkwrite)(struct vm_area_struct *, struct vm_fault *);
 501         int (*access)(struct vm_area_struct *, unsigned long, void*, int, int);
 502
 503 locking rules:
 504                 mmap_sem        PageLocked(page)
 505 open:           yes
 506 close:          yes
 507 fault:          yes             can return with page locked
 508 page_mkwrite:   yes             can return with page locked
 509 access:         yes
 510
 511         ->fault() is called when a previously not present pte is about
 512 to be faulted in. The filesystem must find and return the page associated
 513 with the passed in "pgoff" in the vm_fault structure. If it is possible that
 514 the page may be truncated and/or invalidated, then the filesystem must lock
 515 the page, then ensure it is not already truncated (the page lock will block
 516 subsequent truncate), and then return with VM_FAULT_LOCKED, and the page
 517 locked. The VM will unlock the page.
 518
 519         ->page_mkwrite() is called when a previously read-only pte is
 520 about to become writeable. The filesystem again must ensure that there are
 521 no truncate/invalidate races, and then return with the page locked. If
 522 the page has been truncated, the filesystem should not look up a new page
 523 like the ->fault() handler, but simply return with VM_FAULT_NOPAGE, which
 524 will cause the VM to retry the fault.
 525
 526         ->access() is called when get_user_pages() fails in
 527 acces_process_vm(), typically used to debug a process through
 528 /proc/pid/mem or ptrace.  This function is needed only for
 529 VM_IO | VM_PFNMAP VMAs.
 530
 531 ================================================================================
 532                         Dubious stuff
 533
 534 (if you break something or notice that it is broken and do not fix it yourself
 535 - at least put it here)