It's been tempting to replace this with (L1) cache_line_size(), but
that's really not what the value is about.  It's about coordinating
the condition for the pagelist fragment behavior between the two
sides.  However, the property was not accepted for the upstream DT
binding, so we have to use the firmware's fallback value.
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>