On Fri, Oct 04, 2019 at 10:08:04PM +0200, Uwe Kleine-König wrote:
> On Fri, Oct 04, 2019 at 09:09:51AM +0000, Benjamin GAIGNARD wrote:

> > like for the your patch in IIO trigger regmap_read could only failed
> > if the hardware block is no more clocked and in this case we won't 
> > return of regmap_read.

> I'm not sure this is aligned with how regmap is supposed to be used. I
> think the driver making use of regmap is not supposed to make any
> assumptions about how and when a read or write access can or cannot fail
> and instead is supposed to check all return values. So IMHO the patch
> goes in the right direction.

Yeah, if you're being very good about checking return values you really
should check them all the time in case something comes up - you could
get errors that don't come from the physical access, for example on read
if a register is marked as unreadable, or if physical access is required
but the map has been marked cache only.  That said a lot of people will
just not check anything since it's not common to see errors and errors
that do occur tend to have other quite visible effects like I2C bus
errors or system hangs.