On Mon, Jun 22, 2020 at 03:44:33PM +0300, Kai Vehmanen wrote:
> On Tue, 16 Jun 2020, Kuninori Morimoto wrote:

> > component has io_mutex, but it had been used at
> > snd_soc_component_update_bits_legacy() only which does read and write.

> this is quite interesting. I wonder as regmap does its own locking, should
> io_mutex only be used on the non-regmap paths? The patch looks correct 

It's definitely redundant, though OTOH the redundant locking is easier
to verify.

> though, holding the mutex only in update_bits won't help when write is 
> without locks.

Even with a mutex around the write that'd have potential issues.