Re: [PATCH] driver: bus: simple-pm-bus: Add Freescale i.MX8qm/qxp CSR compatible strings

[Geert Uytterhoeven <geert@linux-m68k.org>]

Hi Krzysztof,

On Sun, Jan 29, 2023 at 11:49 AM Krzysztof Kozlowski
<krzysztof.kozlowski@linaro.org> wrote:
> On 29/01/2023 09:13, Liu Ying wrote:
> > On Thu, 2023-01-26 at 13:45 +0100, Krzysztof Kozlowski wrote:
> >> On 26/01/2023 03:54, Liu Ying wrote:
> >>> On Wed, 2023-01-25 at 10:05 +0100, Geert Uytterhoeven wrote:
> >>>> On Wed, Jan 25, 2023 at 9:31 AM Liu Ying <victor.liu@nxp.com>
> >>>> wrote:
> >>>>> Freescale i.MX8qm/qxp CSR module matches with what the simple
> >>>>> power
> >>>>> managed bus driver does, considering it needs an IPG clock to
> >>>>> be
> >>>>> enabled before accessing it's child devices, the child devices
> >>>>> need
> >>>>> to be populated by the CSR module and the child devices' power
> >>>>> management operations need to be propagated to their parent
> >>>>> devices.
> >>>>> Add the CSR module's compatible strings to
> >>>>> simple_pm_bus_of_match[]
> >>>>> table to support the CSR module.
> >>>>>
> >>>>> Suggested-by: Rob Herring <robh@kernel.org>
> >>>>> Suggested-by: Lee Jones <lee@kernel.org>
> >>>>> Signed-off-by: Liu Ying <victor.liu@nxp.com>

> >>>>> The CSR module's dt-binding documentation can be found at
> >>>>> Documentation/devicetree/bindings/mfd/fsl,imx8qxp-csr.yaml.
> >>>>>
> >>>>> Suggested by Rob and Lee in this thread:

> >>>
> > https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpatchwork.kernel.org%2Fproject%2Flinux-arm-kernel%2Fpatch%2F20221017075702.4182846-1-victor.liu%40nxp.com%2F&data=05%7C01%7Cvictor.liu%40nxp.com%7C87515adc8fc3401f410808daff9b3279%7C686ea1d3bc2b4c6fa92cd99c5c301635%7C0%7C0%7C638103339276325657%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=FFz5gSIPc6vyvb1IJ1Umu62WpzjNLIiQIi2sOA3RQGc%3D&reserved=0
> >>>>>
> >>>>>  drivers/bus/simple-pm-bus.c | 2 ++
> >>>>>  1 file changed, 2 insertions(+)
> >>>>>
> >>>>> diff --git a/drivers/bus/simple-pm-bus.c b/drivers/bus/simple-
> >>>>> pm-
> >>>>> bus.c
> >>>>> index 7afe1947e1c0..4a7575afe6c6 100644
> >>>>> --- a/drivers/bus/simple-pm-bus.c
> >>>>> +++ b/drivers/bus/simple-pm-bus.c
> >>>>> @@ -120,6 +120,8 @@ static const struct of_device_id
> >>>>> simple_pm_bus_of_match[] = {
> >>>>>         { .compatible = "simple-mfd",   .data = ONLY_BUS },
> >>>>>         { .compatible = "isa",          .data = ONLY_BUS },
> >>>>>         { .compatible = "arm,amba-bus", .data = ONLY_BUS },
> >>>>> +       { .compatible = "fsl,imx8qm-lvds-csr", },
> >>>>> +       { .compatible = "fsl,imx8qxp-mipi-lvds-csr", },
> >>>>
> >>>> I did read the thread linked above, and I still think you should
> >>>> just
> >>>> add "simple-pm-bus" to the compatible value in DTS, so no driver
> >>>> change
> >>>> is needed, cfr.
> >>>> Documentation/devicetree/bindings/bus/renesas,bsc.yaml.
> >>
> >> I don't think we want to start putting specific compatibles here. We
> >> don't do it for simple-mfd, syscon and simple-bus, so neither should
> >> we
> >> do it here.
> >>
> >>> This means that i.MX8qm/qxp CSR module dt-binding documentation
> >>> needs
> >>> to be changed.  I'd like to know how Rob and Krzysztof think about
> >>> that.
> >>
> >> The fsl,imx8qxp-csr.yaml bindings are broken anyway... You have
> >> device
> >> specific bindings for non-simple device but use simple-mfd. You
> >> cannot.
> >> simple-mfd means it is simple and none of the resources are needed
> >> for
> >> children, but that binding contradicts it.
> >>
> >> Now you kind of try to extend it even more make it more and more
> >> broken.
> >>
> >> Rework the bindings keeping them backwards compatible. The
> >> combination
> >> with simple-mfd should be deprecated and you can add whatever is
> >> needed
> >> for a proper setup.
> >
> > I did try to rework the bindings and make the combination with simple-
> > mfd deprecated. However, it reminds me the problem that "simple-pm-bus"
> > and "syscon" can not be in compatible string at the same time,
> > otherwise, nodename should match '^syscon@[0-9a-f]+$' and '^bus@[0-9a-
> > f]+$' at the same time. I mentioned the problem in the same thread[1]
> > where Rob and Lee suggest to go with this patch. "syscon" is needed
> > since i.MX8qxp MIPI DSI/LVDS combo PHY node references the CSR module
> > through a phandle, so dropping/deprecating "syscon" is a no-go.
> >
> > Also, as Rob mentioned in [1] "if register space is all mixed together,
> > then it is the former and an MFD", I think the CSR module should fall
> > into the simple-mfd category.
>
> You are now mixing MFD with simple-mfd. If you have clocks there or any
> other resources, it's not simple-mfd anymore.

I beg to differ (but not w.r.t. the other resources): if any "glue" device
between parent and child hierarchies does not call pm_runtime_enable(),
Runtime PM (which is a Linux thing, not a DT thing) for the children
may not work correctly, as it won't propagate correctly to the parent.
So IMHO this is something to fix in Linux, not in DT.

> > Take i.MX8qxp MIPI DSI/LVDS CSR module as
> > an example, child device pxl2dpi register offset is 0x40, while child
> > device ldb register offsets are 0x20 and 0xe0.
> >
> > Geert, Krzysztof, can you please consider to keep this patch as-is,
> > since it seems that there is no other option?
>
> There are other options, why do you say there is no? Making it proper
> binding/driver for its children without abusing simple bindings. Simple
> bindings are for simple cases and this turns out not the simple case.

Or drop the ".data = ONLY_BUS" for "simple-mfd"?
(and treat "simple-bus" the same as "simple-pm-bus", too ;-)

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

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