[PATCH v3] arm64: dts: rockchip: change pinctrl for pcie2x1l2 for Radxa ROCK 5A

[Heiko Stübner]

Am Mittwoch, 2. Oktober 2024, 13:42:05 CEST schrieb Jonas Karlman:
> On 2024-10-02 13:21, Heiko Stübner wrote:
> > Am Mittwoch, 2. Oktober 2024, 12:50:29 CEST schrieb Jonas Karlman:
> >> Hi,
> >>
> >> On 2024-10-02 01:50, FUKAUMI Naoki wrote:
> >>> for pcie2x1l2, only pcie20x1_2_perstn_m0 is required, and its function
> >>> needs to be GPIO to avoid freeze at "pci enum" without PCIe device on
> >>> u-boot.
> >>>
> >>> change pinctrl definitions for pcie2x1l2. no functional change is
> >>> intended on Linux kernel.
> >>
> >> After the split and addition of pcie2_reset I think this patch is no
> >> longer needed? The issue this patch tried to fix was already
> >> fixed/changed in "arm64: dts: rockchip: Split up RK3588's PCIe pinctrls".
> >>
> >> Looks like this now just rename pcie2_reset to pcie20x1_2_perstn_m0?
> > 
> > and removes the other pinctrl states clkreqn and waken .
> 
> This is not something that I think should have been done in the first
> place, the pins/signals exists in hw and schematics, however software
> is not using these signals.
> 
> > 
> > In a previous version they mentioned that this somehow affects u-boot.
> 
> The issue with U-Boot is that for pcie the pin used for reset-gpios must
> use gpio func pinconf, or pci enumerating in U-Boot will freeze the
> board.
> 
> "arm64: dts: rockchip: Split up RK3588's PCIe pinctrls" already changed
> to use the gpio func for the perstn pin, so the U-Boot issue should
> already have been resolved with that patch.
> 
> For Linux the pins gpio func is implicitly configured when the pcie
> driver request the reset-gpios pin. In U-Boot there exists some pinctrl
> overrides to solve freeze issues, those should also be sent upstream ;-)

ok, so I'll just drop this patch for now.

If some other use comes up, please resubmit.

Thanks
Heiko