On Mon, 1 Apr 2019 13:40:33 -0700
Tony Lindgren <tony@atomide.com> wrote:

> * Andreas Kemnade <andreas@kemnade.info> [190401 18:39]:
> > On Mon, 1 Apr 2019 10:52:45 -0700
> > Tony Lindgren <tony@atomide.com> wrote:
> >   
> > > Hi,
> > > 
> > > * Keerthy <j-keerthy@ti.com> [190322 17:16]:  
> > > > +static int am43xx_check_off_mode_enable(void)
> > > > +{
> > > > +	/*
> > > > +	 * Check for am437x-sk-evm which due to HW design cannot support
> > > > +	 * this mode reliably.
> > > > +	 */
> > > > +	if (of_machine_is_compatible("ti,am437x-sk-evm") && enable_off_mode) {
> > > > +		pr_warn("WARNING: This platform does not support off-mode, entering DeepSleep suspend.\n");
> > > > +		return 0;
> > > > +	}
> > > > +
> > > > +	return enable_off_mode;
> > > > +}    
> > > 
> > > Considering off-mode suspend depends on how the board is
> > > wired for various things such as memory, PMIC and the related
> > > signal lines, I agree using the machine compatible is the best
> > > check we can do here.
> > > 
> > > But since the device can hang during suspend unless things are
> > > configured right for the board, I suggest you rather list allowed
> > > boards here that are known to work with off-mode.
> > >   
> > Could we somehow describe this property of the hardware
> > (is-offmode-capable or is-wired-for-offmode) as a separate devicetree
> > property of the soc?
> > 
> > In mmc we have for example "cap-power-off-card" for
> > indicating some is-wired-suitable-for thing.  
> 
> And we also have "regulator-off-in-suspend".
> 
> How about "soc-off-in-suspend" for the generic name?
> 
Well, remember my "omap3: give off mode enable a more prominent place"
maybe we can use the same  capability property for both proposes.

I was a bit unsure about it, so I did not continue with it yet

Regards,
Andreas