cm_split_idlest doesn't take the CM base offset into account right now,
causing it to read reserved registers which show idlestatus as active
always. This will cause the wait_module_ready functionality to be
effectively an expensive NOP, which will cause problems if the
module hasn't really activated during its execution. Fix by adding
the CM offset into the calculation so the wait_module_ready will
access correct register.
Signed-off-by: Tero Kristo <t-kristo@ti.com>
Reported-by: Filip Matijevic <filip.matijevic.pz@gmail.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>