Merge branch 'work.dcache' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs

Pull misc dcache updates from Al Viro:
 "Most of this pile is putting name length into struct name_snapshot and
  making use of it.

  The beginning of this series ("ovl_lookup_real_one(): don't bother
  with strlen()") ought to have been split in two (separate switch of
  name_snapshot to struct qstr from overlayfs reaping the trivial
  benefits of that), but I wanted to avoid a rebase - by the time I'd
  spotted that it was (a) in -next and (b) close to 5.1-final ;-/"

* 'work.dcache' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  audit_compare_dname_path(): switch to const struct qstr *
  audit_update_watch(): switch to const struct qstr *
  inotify_handle_event(): don't bother with strlen()
  fsnotify: switch send_to_group() and ->handle_event to const struct qstr *
  fsnotify(): switch to passing const struct qstr * for file_name
  switch fsnotify_move() to passing const struct qstr * for old_name
  ovl_lookup_real_one(): don't bother with strlen()
  sysv: bury the broken "quietly truncate the long filenames" logics
  nsfs: unobfuscate
  unexport d_alloc_pseudo()

Documentation/filesystems/porting

@@ -668,3 +668,8 @@ in your dentry operations instead.
 	DCACHE_RCUACCESS is gone; having an RCU delay on dentry freeing is the
 	default.  DCACHE_NORCU opts out, and only d_alloc_pseudo() has any
 	business doing so.
+--
+[mandatory]
+	d_alloc_pseudo() is internal-only; uses outside of alloc_file_pseudo() are
+	very suspect (and won't work in modules).  Such uses are very likely to
+	be misspelled d_alloc_anon().