This is now very standard behaviour for CODECs so shouldn't be device
specific and we shouldn't really be trying to peer into the register
cache from atomic context anyway.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>