Hi,
On Thu, Dec 15, 2022 at 08:39:47PM +0100, Marek Vasut wrote:
> On 12/15/22 19:06, Sebastian Reichel wrote:
> > On Sun, Dec 11, 2022 at 09:51:24PM +0100, Marek Vasut wrote:
> > > The clock {} subnode seems like it is describing an always-on clock
> > > generated by the PMIC. This should rather be modeled by consumer of
> > > the clock taking phandle to the RTC node itself, since it already
> > > does have clock-cells and all. Since there are no users of the clock
> > > subnode in tree anyway, mark it as deprecated to avoid proliferation
> > > of this approach.
> > >
> > > Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxx>
> > > Signed-off-by: Marek Vasut <marex@xxxxxxx>
> > > ---
> > > Cc: Alessandro Zummo <a.zummo@xxxxxxxxxxxx>
> > > Cc: Alexandre Belloni <alexandre.belloni@xxxxxxxxxxx>
> > > Cc: Rob Herring <robh+dt@xxxxxxxxxx>
> > > Cc: Krzysztof Kozlowski <krzysztof.kozlowski+dt@xxxxxxxxxx>
> > > Cc: linux-rtc@xxxxxxxxxxxxxxx
> > > To: devicetree@xxxxxxxxxxxxxxx
> > > ---
> > > V2: - Add AB from Krzysztof
> > > V3: - No change
> > > ---
> >
> > I just noticed this by accident. Basically everything in the patch
> > description is wrong:
> >
> > 1. There is a in-tree user: arch/arm/boot/dts/imx6dl-qmx6.dtsi
>
> Sorry, I missed this one.
>
> > 2. The PMIC has nothing to do with this
>
> In [3] the commit message claims the PMIC supplies 32kHz clock to i.MX6 CKIL,
> which per IMX6DQRM rev.6 Table 18-3 row SNVS indirectly supplies SNVS RTC.
> This reminded me of commit:
The word PMIC is not mentioned once in [3]. PMIC is not involved.
The QMX6 32khz chain is like this:
32kHz crystal -> m41t62 crystal input
m41t62 clock output -> i.MX6 CKIL
> 9509593f327ac ("arm64: dts: imx8mm: Model PMIC to SNVS RTC clock path on
> Data Modul i.MX8M Mini eDM SBC")
>
> which solves exactly the same problem, system hangs when 32 kHz clock are
> stopped, except this time on i.MX8MM, clock are generated by PMIC on I2C
> (notice how the PMIC is referenced directly) and the clock are supplied to
> the SVNS RTC XTal terminals.
>
> I wonder if this could be reused on the QMX6 board too?
IIRC On i.MX6 referencing the I2C connected RTC results in boot
hanging forever when trying to get the ckil clock in
imx6q_clocks_init. At least it used to be the case when I was
working on this - I no longer have access to the boards. Of course
properly referencing the RTC clock was the first route I tried.
-- Sebastian
> > 3. Directly referencing the RTC does not work, since that introduces
> > an unsolvable dependency loop on QMX6. This was the solution accepted
> > by Rob and Saravana:
> >
> > [v1] https://lore.kernel.org/lkml/20210222171247.97609-1-sebastian.reichel@xxxxxxxxxxxxx/
> > [v2] https://lore.kernel.org/all/20210428222953.235280-1-sebastian.reichel@xxxxxxxxxxxxx/
>
> [3] https://lore.kernel.org/linux-clk/20191108170135.9053-1-sebastian.reichel@xxxxxxxxxxxxx/
Attachment:
signature.asc
Description: PGP signature
