Re: [PATCH] phy: qcom: qmp-pcie: drop bogus x1e80100 qref supply

[Johan Hovold <johan@kernel.org>]

On Tue, Apr 29, 2025 at 04:24:19PM +0800, Qiang Yu wrote:
> 
> On 4/29/2025 3:54 PM, Johan Hovold wrote:
> > The PCIe PHYs on x1e80100 do not a have a qref supply so stop requesting
> > one. This also avoids the follow warning at boot:
> >
> > 	qcom-qmp-pcie-phy 1be0000.phy: supply vdda-qref not found, using dummy regulator
> >
> > Fixes: e961ec81a39b ("phy: qcom: qmp: Add phy register and clk setting for x1e80100 PCIe3")
> > Cc: Qiang Yu <quic_qianyu@quicinc.com>
> > Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
> > ---

> We have QREF for each PCIe port on the X1E80100, all of which consume 
> the regulator L3J. Although the PCIe PHY uses QREF indirectly, this 
> creates a dependency, right? 

The PHY binding should describe the direct dependencies for the PHY, so
the addition of qref for sm8550/sm8650 was probably also a mistake.

From what I could tell there is not even a one-to-one mapping of qref
supplies to PCIe ports, but perhaps you can provide more details on how
this fits together here?

> If PCIe doesn't vote for it, how can the 
> PMIC driver decide when to disable L3J during system suspend or runtime 
> suspend? Is there a chance that L3J could be disabled while PCIe still 
> requires it?

If the QREF supplies can be turned off, you may need to mark them as
always-on until things are described properly. But whether that's needed
is not even clear at this point:

	https://lore.kernel.org/lkml/17a1a4d9-fdc5-477a-bf4e-91cae5a62479@oss.qualcomm.com/

Johan