the way it abuses ->d_fsdata still needs to be killed, but that's
a separate story.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>