Re: [PATCH v1] ARM: dts: exynos: Add proper regulator states for suspend-to-mem for odroid-u3

From: Krzysztof Kozlowski
Date: Mon Mar 25 2019 - 08:56:47 EST

On Mon, 25 Mar 2019 at 13:46, Krzysztof Kozlowski <krzk@xxxxxxxxxx> wrote:
>
> On Sun, 24 Mar 2019 at 09:33, Anand Moon <linux.amoon@xxxxxxxxx> wrote:
> >
> > Add suspend-to-mem node to regulator core to be enabled or disabled
> > during system suspend and also support changing the regulator operating
> > mode during runtime and when the system enter sleep mode (stand by mode).
> >
> > Cc: Marek Szyprowski <m.szyprowski@xxxxxxxxxxx>
> > Cc: Krzysztof Kozlowski <krzk@xxxxxxxxxx>
> > Cc: Chanwoo Choi <cw00.choi@xxxxxxxxxxx>
> > Signed-off-by: Anand Moon <linux.amoon@xxxxxxxxx>
> > ---
> > Current patch:
> >
> > Note: Both microSD and eMMC suspend resume works this changes at my end.
> >
> > regulator-off-in-suspend:
> > set the regulator node into suspend state i.e. standby mode during suspend
> > operation.
> >
> > Current changes are based on
> > [0] https://www.kernel.org/doc/Documentation/devicetree/bindings/regulator/max77686.txt
> >
> > Regulators which can be turned off during system suspend:
> > -LDOn : 2, 6-8, 10-12, 14-16,
> > -BUCKn : 1-4.
> > Use standard regulator bindings for it ('regulator-off-in-suspend').
> >
> > drop the suspend off binding which are not supported by the driver.
> >
> > RFC version
> > [1] https://patchwork.kernel.org/patch/10810909/
> > These changes had some problem with eMMC not entering into suspend mode.
> > with some miss configuration in regulator-off-in-suspend mode.
> >
> > Changes from previos patch.
> > [2] https://patchwork.kernel.org/patch/10712549/
> >
> > Set all the non used regulator in suspend-odd state
> > LD02, LD03, LD05, LD06, LD07, LD011, LD013, LDO14, LD016
> >
> > BUCK5, BUCK6, BUCK7 and not confirable as per driver max77686-regulator
> > ---
> > .../boot/dts/exynos4412-odroid-common.dtsi | 39 +++++++++++++++++++
> > 1 file changed, 39 insertions(+)
> >
> > diff --git a/arch/arm/boot/dts/exynos4412-odroid-common.dtsi b/arch/arm/boot/dts/exynos4412-odroid-common.dtsi
> > index 08d3a0a7b4eb..375156ad5454 100644
> > --- a/arch/arm/boot/dts/exynos4412-odroid-common.dtsi
> > +++ b/arch/arm/boot/dts/exynos4412-odroid-common.dtsi
> > @@ -288,6 +288,9 @@
> > regulator-min-microvolt = <1800000>;
> > regulator-max-microvolt = <1800000>;
> > regulator-always-on;
> > + regulator-state-mem {
> > + regulator-off-in-suspend;
>
> Please maintain proper versioning of patches.
> First patch sent to lists should be either RFC or v1. Second
> release/submission is then always v2. Third v3.
>
> This is third or fourth submission but you marked it as v1. This makes
> it very difficult to discuss and reference previous versions.
>
> The commit message did not change since beginning (first version). I
> asked twice that you need to explain exactly why you put the the
> regulator to off or on state in suspend. Why?
> Because:
> 1. This change looks without justification - once you put on, then you
> put off, now again on,
> 2. Anyone reading the code later must know the rationale why this was done,
> 3. I am not quite sure whether this is good setting so I would be
> happy to be convinced.
>
> How to provide such explanation? The best in commit message. Sometimes
> in the comment in the code, depends.
> How such explanation could look like? For example like this:
> f8f3b7fc21b1 ("ARM: dts: exynos: Fix regulators configuration on Peach
> Pi/Pit Chromebooks")
> Marek clearly explained why he put the regulators "always-on", even
> tough we do not know everything about this. More over, he mentions
> that this fixes specific issue.
>
> Summarizing, please answer:
> 1. Why this is made off-in-suspend?
> 2. Why this can be made off-in-suspend?
>
> > + };
> > };
> >
> > ldo3_reg: LDO3 {
> > @@ -317,6 +320,9 @@
> > regulator-min-microvolt = <1000000>;
> > regulator-max-microvolt = <1000000>;
> > regulator-always-on;
> > + regulator-state-mem {
> > + regulator-off-in-suspend;
> > + };
> > };
> >
> > ldo7_reg: LDO7 {
> > @@ -324,18 +330,27 @@
> > regulator-min-microvolt = <1000000>;
> > regulator-max-microvolt = <1000000>;
> > regulator-always-on;
> > + regulator-state-mem {
> > + regulator-off-in-suspend;
> > + };
> > };
> >
> > ldo8_reg: LDO8 {
> > regulator-name = "VDD10_HDMI_1.0V";
> > regulator-min-microvolt = <1000000>;
> > regulator-max-microvolt = <1000000>;
> > + regulator-state-mem {
> > + regulator-off-in-suspend;
> > + };
> > };
> >
> > ldo10_reg: LDO10 {
> > regulator-name = "VDDQ_MIPIHSI_1.8V";
> > regulator-min-microvolt = <1800000>;
> > regulator-max-microvolt = <1800000>;
> > + regulator-state-mem {
> > + regulator-off-in-suspend;
> > + };
> > };
> >
> > ldo11_reg: LDO11 {
> > @@ -343,6 +358,9 @@
> > regulator-min-microvolt = <1800000>;
> > regulator-max-microvolt = <1800000>;
> > regulator-always-on;
> > + regulator-state-mem {
> > + regulator-off-in-suspend;
> > + };
> > };
> >
> > ldo12_reg: LDO12 {
> > @@ -351,6 +369,9 @@
> > regulator-max-microvolt = <3300000>;
> > regulator-always-on;
> > regulator-boot-on;
> > + regulator-state-mem {
> > + regulator-off-in-suspend;
> > + };
> > };
> >
> > ldo13_reg: LDO13 {
> > @@ -367,6 +388,9 @@
> > regulator-max-microvolt = <1800000>;
> > regulator-always-on;
> > regulator-boot-on;
> > + regulator-state-mem {
> > + regulator-off-in-suspend;
> > + };
> > };
> >
> > ldo15_reg: LDO15 {
> > @@ -375,6 +399,9 @@
> > regulator-max-microvolt = <1000000>;
> > regulator-always-on;
> > regulator-boot-on;
> > + regulator-state-mem {
> > + regulator-off-in-suspend;
> > + };
> > };
> >
> > ldo16_reg: LDO16 {
> > @@ -383,6 +410,9 @@
> > regulator-max-microvolt = <1800000>;
> > regulator-always-on;
> > regulator-boot-on;
> > + regulator-state-mem {
> > + regulator-off-in-suspend;
> > + };
> > };
> >
> > ldo20_reg: LDO20 {
> > @@ -421,6 +451,9 @@
> > regulator-max-microvolt = <1100000>;
> > regulator-always-on;
> > regulator-boot-on;
> > + regulator-state-mem {
> > + regulator-off-in-suspend;
> > + };
> > };
>
> I questioned this change two times. You did not answer to my
> question... If you turn memory bus regulator off, how the memory will
> work in Suspend-to-Memory mode? I might be missing here something but
> it just looks suspicious. Maybe the regulator does not supply the
> memory itself (so refresh works even when it is down), just the
> interface? I don't know, it just looks suspicious. I need to see
> proper explanation.
>
> I am sorry but I will not check other hunks in this patch. Please
> provide the answers for all my questions here first.

I started digging around this and datasheet describes the regulator
values in sleep mode. The MIF can be off if C2C is not used. It seems
that the memory refresh is provided by CKEM regulators (which you
wanted to turn off in first version of the patch). I am not happy that
I ask about such information and cannot get it. It is the job of
submitter to provide rationale.

Best regards,
Krzysztof