PCIe missing on RK3399

[Vicente Bergas]

On Wed, Dec 11, 2024 at 4:35 PM Heiko Stübner <heiko at sntech.de> wrote:
>
> Am Mittwoch, 11. Dezember 2024, 16:10:21 CET schrieb Vicente Bergas:
> > On Wed, Dec 11, 2024 at 2:36 PM Heiko Stübner <heiko at sntech.de> wrote:
> > >
> > > Hi Vicente,
> >
> > Hi Heiko,
> > thanks for taking a look at it!
> >
> > > Am Mittwoch, 11. Dezember 2024, 13:55:01 CET schrieb Vicente Bergas:
> > > > i've tested the Linux kernel 6.13-rc1 and rc2 and in both cases PCIe
> > > > is not detected on the RK3399 platform (rk3399-gru-kevin), whereas the
> > > > kernel version 6.12.3 works fine.
> > > >
> > > > 6.13 configuration is based on the same one as 6.12 and there aren't
> > > > any significant PCI-related differences.
> > > >
> > > > The messages from dmesg on 6.13 don't show any PCI-related errors.
> > > >
> > > > Does somebody know what is going on?
> > >
> > > so I just booted a rk3399-puma-haikou with a pci-nvme-adapter in the
> > > pcie slot. And I get:
> > >
> > > [    0.000000] Booting Linux on physical CPU 0x0000000000 [0x410fd034]
> > > [    0.000000] Linux version 6.13.0-rc2-00101-g260ae63734ff-dirty (hstuebner at phil) (aarch64-linux-gnu-gcc (Debian 14.2.0-6) 14.2.0, GNU ld (GNU Binutils for Debian) 2.43.1) #1134 SMP PREEMPT Tue Dec 10 21:06:34 CET 2024
> > > ...
> > > [    3.428114] rockchip-pcie f8000000.pcie: host bridge /pcie at f8000000 ranges:
> > > [    3.435978] rockchip-pcie f8000000.pcie:      MEM 0x00fa000000..0x00fbdfffff -> 0x00fa000000
> > > [    3.445478] rockchip-pcie f8000000.pcie:       IO 0x00fbe00000..0x00fbefffff -> 0x00fbe00000
> > > [    3.455298] rockchip-pcie f8000000.pcie: using DT '/pcie at f8000000' for 'ep' GPIO lookup
> > >
> > > So there seems to be not some general failure.
> > >
> > > Does
> > >         # ls /sys/devices/platform/f8000000.pcie
> > > list some "waiting_for_supplies" or something?
> >
> > yes, indeed there is such a file in there:
> >
> > ### 6.13-rc2
> > $ ls /sys/devices/platform/f8000000.pcie
> > power
> > driver_override
> > modalias
> > of_node
> > subsystem
> > supplier:platform:ff720000.gpio
> > supplier:platform:ff770000.syscon:pcie-phy
> > supplier:platform:ff780000.gpio
> > supplier:platform:pinctrl
> > supplier:platform:regulator-pp3300-wifi-bt
> > supplier:platform:regulator-wlan-pd-n
> > uevent
> > waiting_for_supplier
> >
> > $ cat /sys/devices/platform/f8000000.pcie/waiting_for_supplier
> > 1
> >
> > ### 6.12.3
> > $ ls /sys/devices/platform/f8000000.pcie
> > pci0000:00
> > power
> > driver
> > driver_override
> > modalias
> > of_node
> > subsystem
> > supplier:phy:phy-ff770000.syscon:pcie-phy.5
> > supplier:phy:phy-ff770000.syscon:pcie-phy.6
> > supplier:phy:phy-ff770000.syscon:pcie-phy.7
> > supplier:phy:phy-ff770000.syscon:pcie-phy.8
> > supplier:platform:ff770000.syscon:pcie-phy
> > supplier:platform:ff780000.gpio
> > supplier:platform:pinctrl
> > supplier:platform:pp3300-wifi-bt
> > supplier:platform:pp900-ap
> > supplier:platform:wlan-pd-n
> > supplier:regulator:regulator.17
> > supplier:regulator:regulator.23
> > supplier:regulator:regulator.25
> > uevent
> >
> > What does that mean?
>
> waiting_for_supplier means that some supplier has not yet probed
> and thus the driver also cannot probe yet.
>
> But your 6.13-rc2 supplier list does look way too short. In my boot on
> rk3399-puma-haikou above, I got:
>
> $ ls /sys/devices/platform/f8000000.pcie
> driver
> driver_override
> modalias
> of_node
> pci0000:00
> power
> subsystem
> supplier:phy:phy-ff770000.syscon:pcie-phy.2
> supplier:phy:phy-ff770000.syscon:pcie-phy.3
> supplier:phy:phy-ff770000.syscon:pcie-phy.4
> supplier:phy:phy-ff770000.syscon:pcie-phy.5
> supplier:platform:ff770000.syscon:pcie-phy
> supplier:platform:ff790000.gpio
> supplier:platform:pinctrl
> supplier:platform:regulator-dc-12v
> supplier:platform:regulator-vcc3v3-baseboard
> supplier:platform:regulator-vcca-0v9
> supplier:platform:regulator-vcca-1v8
> supplier:regulator:regulator.10
> supplier:regulator:regulator.11
> supplier:regulator:regulator.29
> supplier:regulator:regulator.3
> uevent

Just tested 6.13-rc2 with the DTB from 6.12.3 and it works fine. So,
it is a DT issue.
I suspect this commit may be the root cause:
5c96e63301978f4657c9082c55a066763c8db7b1
arm64: dts: rockchip: adapt regulator nodenames to preferred form