The .set_suspend_enable() and .set_suspend_disable() methods are not
supposed to immediately change the regulator state but just indicated
if the regulator should be enabled or disabled when standby mode is
entered (by a hardware signal).

However currently they set control the SEL bits in the DVC registers,
which causes the voltage to change to immediately between the "A"
(normal) and "B" (standby) values as programmed and does nothing for
the enable state...

This means that "regulator-on-in-suspend" does not work (the regulator
is switched off when the PMIC enters standby mode on the hardware
signal) and, potentially, depending on the A and B voltage
configurations the voltage could be incorrectly changed *before*
actually entering suspend.

The right bit to use for the functionality is the "CONF" bit in the
"CONT" register.
The detailed register description says "Sequencer target state"
for this bit which is not very clear but the functional description
is clearer.

>From 5.1.5 System Enable:

	De-asserting SYS_EN (changing from active to passive state)
	clears control SYSTEM_EN  which triggers a power down sequence
	into hibernate/standby mode
	...
	With the exception of supplies that have the xxxx_CONF control
	bit asserted, all regulators in power domains POWER1, POWER, and
	SYSTEM are sequentially disabled in reverse order.
	Regulators with the <x>_CONF bit set remain on but change the
	active voltage controlregisters from V<x>_A to V<x>_B
	(if V<x>_B is notalready selected).
Signed-off-by: Martin Fuzzey <martin.fuzzey@flowbird.group>
Link: https://lore.kernel.org/r/1584461691-14344-1-git-send-email-martin.fuzzey@flowbird.groupSigned-off-by: Mark Brown <broonie@kernel.org>