Re: [PATCH v2] arm64: dts: rockchip: Add missing pinctrl for PCIe30x4 node

[Alex Bee]

Am 11.07.24 um 13:38 schrieb Chen-Yu Tsai:
> On Thu, Jul 11, 2024 at 7:13 PM Alex Bee <knaerzche@xxxxxxxxx> wrote:
> > Am 11.07.24 um 11:17 schrieb Chen-Yu Tsai:
> > > On Thu, Jul 11, 2024 at 4:44 PM Jonas Karlman <jonas@xxxxxxxxx> wrote:
> > > > Hi Anand,
> > > >
> > > > On 2024-07-11 08:09, Anand Moon wrote:
> > > > > Add missing pinctrl settings for PCIe 3.0 x4 clock request and wake
> > > > > signals.Each component of PCIe communication have the following control
> > > > > signals: PERST, WAKE, CLKREQ, and REFCLK. These signals work to generate
> > > > > high-speed signals and communicate with other PCIe devices.
> > > > > Used by root complex to endpoint depending on the power state.
> > > > >
> > > > > PERST is referred to as a fundamental reset. PERST should be held low
> > > > > until all the power rails in the system and the reference clock are stable.
> > > > > A transition from low to high in this signal usually indicates the
> > > > > beginning of link initialization.
> > > > >
> > > > > WAKE signal is an active-low signal that is used to return the PCIe
> > > > > interface to an active state when in a low-power state.
> > > > >
> > > > > CLKREQ signal is also an active-low signal and is used to request the
> > > > > reference clock.
> > > > >
> > > > > Signed-off-by: Anand Moon <linux.amoon@xxxxxxxxx>
> > > > > ---
> > > > > V2: Update the commit messge to describe the changs.
> > > > >       use pinctl group as its pre define in pinctl dtsi
> > > > > ---
> > > > >    arch/arm64/boot/dts/rockchip/rk3588-rock-5b.dts | 6 +-----
> > > > >    1 file changed, 1 insertion(+), 5 deletions(-)
> > > > >
> > > > > diff --git a/arch/arm64/boot/dts/rockchip/rk3588-rock-5b.dts b/arch/arm64/boot/dts/rockchip/rk3588-rock-5b.dts
> > > > > index 2e7512676b7e..ab3a20986c6a 100644
> > > > > --- a/arch/arm64/boot/dts/rockchip/rk3588-rock-5b.dts
> > > > > +++ b/arch/arm64/boot/dts/rockchip/rk3588-rock-5b.dts
> > > > > @@ -301,7 +301,7 @@ &pcie30phy {
> > > > >
> > > > >    &pcie3x4 {
> > > > >         pinctrl-names = "default";
> > > > > -     pinctrl-0 = <&pcie3_rst>;
> > > > > +     pinctrl-0 = <&pcie30x4m1_pins>;
> > > > Use of the existing pcie30x4m1_pins group may not be fully accurate for
> > > > the PERST pin. The use of reset-gpios indicate that the PERST pin is
> > > > used with GPIO function and the driver will implicitly change the
> > > > function from perstn_m1 to GPIO. So this may not be best representation
> > > > of the hw, hence my initial suggestion, something like:
> > > >
> > > >           pcie30x4_pins: pcie30x4-pins {
> > > >                   rockchip,pins =
> > > >                           <4 RK_PB4 4 &pcfg_pull_none>,
> > > >                           <4 RK_PB6 RK_FUNC_GPIO &pcfg_pull_none>,
> > > >                           <4 RK_PB5 4 &pcfg_pull_none>;
> > > >           };
> > > >
> > > > Similar change should probably also be done for pcie2x1l0 and pcie2x1l2,
> > > > not just pcie3x4.
> > > Can we consider implementing strict mode in the pinctrl driver so we don't
> > > have to keep doing this GPIO + pinmux dance?
> > This is not about drivers, this is DT which is independent from drivers.
> > Jonas and I had discussion recently on u-boot mailing list with somebody
> > using Open/FreeBSD about exact that topic. Pinctrl mux settings should be
> > very explicit, even if it is not required by the linux-driver.
> The hardware also prevents you from using GPIOs with any other mux setting.
> That is already implied by the compatible string and doesn't need to be
> spelled out. The strict driver mode merely enforces this hardware limitation
> without the very explicit setting. One could also say that the driver is
> wrong to support conflicting settings.
>
>
> ChenYu
"The" driver is in fact not correct here, since it's two actually two
drivers: pinctrl and gpio, two subsystems and separate HW blocks (not very
clear, but it's at least configured in different mmio regions). So with
what you are suggesting you want one driver to implicitly configure the
other and expect every OS to have such methods? Broad ...

Alex

> > Alex
> >
> > > ChenYu
> > >
> > >
> > > > Regards,
> > > > Jonas
> > > >
> > > > >         reset-gpios = <&gpio4 RK_PB6 GPIO_ACTIVE_HIGH>;
> > > > >         vpcie3v3-supply = <&vcc3v3_pcie30>;
> > > > >         status = "okay";
> > > > > @@ -341,10 +341,6 @@ pcie2_2_rst: pcie2-2-rst {
> > > > >         };
> > > > >
> > > > >         pcie3 {
> > > > > -             pcie3_rst: pcie3-rst {
> > > > > -                     rockchip,pins = <4 RK_PB6 RK_FUNC_GPIO &pcfg_pull_none>;
> > > > > -             };
> > > > > -
> > > > >                 pcie3_vcc3v3_en: pcie3-vcc3v3-en {
> > > > >                         rockchip,pins = <1 RK_PA4 RK_FUNC_GPIO &pcfg_pull_none>;
> > > > >                 };
> > > > >
> > > > > base-commit: 34afb82a3c67f869267a26f593b6f8fc6bf35905
> > > _______________________________________________
> > > Linux-rockchip mailing list
> > > Linux-rockchip@xxxxxxxxxxxxxxxxxxx
> > > http://lists.infradead.org/mailman/listinfo/linux-rockchip