Re: [RESEND PATCH] arm64: dts: rockchip: Add PCIe pinctrls to Turing RK1

[Heiko Stübner]

Am Mittwoch, 6. Dezember 2023, 19:26:40 CET schrieb Sam Edwards:
> 
> On 12/6/23 07:55, Heiko Stübner wrote:
> > Am Dienstag, 5. Dezember 2023, 21:28:59 CET schrieb Sam Edwards:
> >> The RK3588 PCIe 3.0 controller seems to have unpredictable behavior when
> >> no CLKREQ/PERST/WAKE pins are configured in the pinmux. In particular, it
> >> will sometimes (varying between specific RK3588 chips, not over time) shut
> >> off the DBI block, and reads to this range will instead stall
> >> indefinitely.
> >>
> >> When this happens, it will prevent Linux from booting altogether. The
> >> PCIe driver will stall the CPU core once it attempts to read the version
> >> information from the DBI range.
> >>
> >> Fix this boot hang by adding the correct pinctrl configuration to the
> >> PCIe 3.0 device node, which is the proper thing to do anyway. While
> >> we're at it, also add the necessary configuration to the PCIe 2.0 node,
> >> which may or may not fix the equivalent problem over there -- but is the
> >> proper thing to do anyway. :)
> >>
> >> Fixes: 2806a69f3fef6 ("arm64: dts: rockchip: Add Turing RK1 SoM support")
> >> Signed-off-by: Sam Edwards <CFSworks@xxxxxxxxx>
> >> ---
> >>   .../arm64/boot/dts/rockchip/rk3588-turing-rk1.dtsi | 14 ++------------
> >>   1 file changed, 2 insertions(+), 12 deletions(-)
> >>
> >> diff --git a/arch/arm64/boot/dts/rockchip/rk3588-turing-rk1.dtsi b/arch/arm64/boot/dts/rockchip/rk3588-turing-rk1.dtsi
> >> index 9570b34aca2e..129f14dbd42f 100644
> >> --- a/arch/arm64/boot/dts/rockchip/rk3588-turing-rk1.dtsi
> >> +++ b/arch/arm64/boot/dts/rockchip/rk3588-turing-rk1.dtsi
> >> @@ -214,7 +214,7 @@ rgmii_phy: ethernet-phy@1 {
> >>   &pcie2x1l1 {
> >>   	linux,pci-domain = <1>;
> >>   	pinctrl-names = "default";
> >> -	pinctrl-0 = <&pcie2_reset>;
> >> +	pinctrl-0 = <&pcie30x1m1_pins>;
> >>   	reset-gpios = <&gpio4 RK_PA2 GPIO_ACTIVE_HIGH>;
> >>   	status = "okay";
> >>   };
> >> @@ -226,7 +226,7 @@ &pcie30phy {
> >>   &pcie3x4 {
> >>   	linux,pci-domain = <0>;
> >>   	pinctrl-names = "default";
> >> -	pinctrl-0 = <&pcie3_reset>;
> >> +	pinctrl-0 = <&pcie30x4m1_pins>;
> 
> Hi Heiko,
> 
> > 
> > also, why are you throwing out the pinctrl for the reset pin.
> > That seems not related to the regular pins and you could instead just do
> > 
> > +	pinctrl-0 = <&pcie30x4m1_pins>, <&pcie3_reset>;
> 
> The pcie30x4m1_pins def does include pinmuxing `4 RK_PB6` to the DW 
> controller already. The v2 patch should probably instead remove the 
> reset-gpios property, since an out-of-band GPIO reset is not needed when 
> the controller can do it.

yep, also because of the reset-gpios the pinctrl/gpio driver will mux the
pin to gpio function even though the pinctrl would've set if the pcie-
function before that.

So I'm really in favor of not mixing the two concepts :-)
When setting the pins, the reset-gpio should be gone and vice-versa.

> I'm still looking into the story with the PCIe 2.0 pins, since 2.0x1's 
> PERST# is definitely 4A2. I'll ask around and try to find out where the 
> corresponding CLKREQ# is going.

Yeah, I tried reading up in the TRM but it was really hard following
which pin-group actually goes to which controller and the naming
definitly does not help :-) .


Heiko



> >>   	reset-gpios = <&gpio4 RK_PB6 GPIO_ACTIVE_HIGH>;
> >>   	vpcie3v3-supply = <&vcc3v3_pcie30>;
> >>   	status = "okay";
> >> @@ -245,17 +245,7 @@ hym8563_int: hym8563-int {
> >>   		};
> >>   	};
> >>   
> >> -	pcie2 {
> >> -		pcie2_reset: pcie2-reset {
> >> -			rockchip,pins = <4 RK_PA2 RK_FUNC_GPIO &pcfg_pull_none>;
> >> -		};
> >> -	};
> >> -
> >>   	pcie3 {
> >> -		pcie3_reset: pcie3-reset {
> >> -			rockchip,pins = <4 RK_PB6 RK_FUNC_GPIO &pcfg_pull_none>;
> >> -		};
> >> -
> >>   		vcc3v3_pcie30_en: pcie3-reg {
> >>   			rockchip,pins = <2 RK_PC5 RK_FUNC_GPIO &pcfg_pull_none>;
> >>   		};
> >>
> > 
> > 
> > 
> > 
>