On Tue, Jul 07, 2020 at 12:53:38PM +0200, Adrian Fiergolski wrote:
> On 06.07.2020 18:18, Mark Brown wrote:
> > On Mon, Jul 06, 2020 at 11:22:43AM +0200, Adrian Fiergolski wrote:

> > this.  Perhaps this needs to be a library for devices that can daisy
> > chain?  It does feel like the instances should be aware of each other
> > since half the point with building the hardware like this is that it
> > enables operations on multiple devices to happen in sync.

> Well, I assume that one can connect completely different SPI devices on
> a single daisy chain. In order to address all devices in a single
> access, a controller would need to delay a transaction, wait for a
> certain moment (how long?), synchronise and then transfer the message. I
> think it adds unnecessary complexity, as MHO many engineers' issue at

I'd imagine something more like regmap's explicit management with
pausing and unpausing transactions (indeed regmap might be a good
mechanism for this for devices with a register map) would not be that
cumbersome to use.  Pause, set the desired target state and then sync
that state out to the hardware.  This isn't something that can sensibly
be done at the controller level, the devices do need to be aware of
what's going on, which is why I'm saying that there's a need for the
devices to be aware of things.