Re: [PATCH v2 1/8] PCI: rcar-gen2: Add support for clocks

[Geert Uytterhoeven]

Hi Hervé,

On Thu, Apr 14, 2022 at 1:29 PM Herve Codina <herve.codina@xxxxxxxxxxx> wrote:
> On Thu, 14 Apr 2022 10:45:54 +0200
> Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
> > On Thu, Apr 14, 2022 at 9:40 AM Herve Codina <herve.codina@xxxxxxxxxxx> wrote:
> > > The PCI rcar-gen2 does not call any clk_prepare_enable().
> >
> > Correct, this driver manages the clocks indirectly through Runtime PM.
> >
> > > This lead to an access failure when the driver tries to access
> > > the IP (at least on a RZ/N1D platform).
> >
> > I expect adding
> >
> >     power-domans = <&sysctrl>;
> >
> > to the pci_usb node makes this patch redundant.
>
> Seems not enough.
> I tried what you suggest :
>  - Added 'power-domains = <&systrl>;' to the pci_usb node
>  - Added missing '#power-domain-cells = <0>;' to sysctrl node
>  - Reverted my patch.
>
> The system crashed at boot:

> [    0.832958] Unhandled fault: external abort on non-linefetch (0x1008) at 0x90b5f848

That's indeed a typical symptom of accessing a module's registers
while the module's clock is disabled.

> I also added a trace printk in r9a06g032-clocks.c and
> r9a06g032_attach_dev() was never called.
>
> Did I miss to set something ?

Do you have CONFIG_PM and CONFIG_PM_GENERIC_DOMAINS
enabled?
Apparently ARCH_RZN1 does not select these options yet.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds