Re: [PATCH v2 0/4] soc: qcom: Introduce PMIC GLINK

[Bjorn Andersson]

On Tue, Jan 17, 2023 at 11:26:58AM +0200, Dmitry Baryshkov wrote:
> On 17/01/2023 04:58, Bjorn Andersson wrote:
> > On Tue, Jan 17, 2023 at 02:37:27AM +0000, Bryan O'Donoghue wrote:
> > > On 17/01/2023 02:32, Bjorn Andersson wrote:
> > > > On Fri, Jan 13, 2023 at 05:10:17PM +0000, Bryan O'Donoghue wrote:
> > > > > On 13/01/2023 04:11, Bjorn Andersson wrote:
> > > > > > This implements the base PMIC GLINK driver, a power_supply driver and a
> > > > > > driver for the USB Type-C altmode protocol. This has been tested and
> > > > > > shown to provide battery information, USB Type-C switch and mux requests
> > > > > > and DisplayPort notifications on SC8180X, SC8280XP and SM8350.
> > > > > > 
> > > > > > Bjorn Andersson (4):
> > > > > >      dt-bindings: soc: qcom: Introduce PMIC GLINK binding
> > > > > >      soc: qcom: pmic_glink: Introduce base PMIC GLINK driver
> > > > > >      soc: qcom: pmic_glink: Introduce altmode support
> > > > > >      power: supply: Introduce Qualcomm PMIC GLINK power supply
> > > > > > 
> > > > > >     .../bindings/soc/qcom/qcom,pmic-glink.yaml    |  102 ++
> > > > > >     drivers/power/supply/Kconfig                  |    9 +
> > > > > >     drivers/power/supply/Makefile                 |    1 +
> > > > > >     drivers/power/supply/qcom_battmgr.c           | 1421 +++++++++++++++++
> > > > > >     drivers/soc/qcom/Kconfig                      |   15 +
> > > > > >     drivers/soc/qcom/Makefile                     |    2 +
> > > > > >     drivers/soc/qcom/pmic_glink.c                 |  336 ++++
> > > > > >     drivers/soc/qcom/pmic_glink_altmode.c         |  477 ++++++
> > > > > >     include/linux/soc/qcom/pmic_glink.h           |   32 +
> > > > > >     9 files changed, 2395 insertions(+)
> > > > > >     create mode 100644 Documentation/devicetree/bindings/soc/qcom/qcom,pmic-glink.yaml
> > > > > >     create mode 100644 drivers/power/supply/qcom_battmgr.c
> > > > > >     create mode 100644 drivers/soc/qcom/pmic_glink.c
> > > > > >     create mode 100644 drivers/soc/qcom/pmic_glink_altmode.c
> > > > > >     create mode 100644 include/linux/soc/qcom/pmic_glink.h
> > > > > > 
> > > > > 
> > > > > How does the USB PHY and a USB redriver fit into this ?
> > > > > 
> > > > > Is the host supposed to manage both/neither ? Is the DSP responsible for
> > > > > configuring the PHY lanes and the turnaround on orientation switch ?
> > > > > 
> > > > 
> > > > As indicated above, the firmware deals with battery management and USB
> > > > Type-C handling.
> > > > 
> > > > The battery/power management is handled by the battmgr implementation,
> > > > exposing the various properties through a set of power_supply objects.
> > > > 
> > > > The USB Type-C handling comes in two forms. The "altmode" protocol
> > > > handles DisplayPort notifications - plug detect, orientation and mode
> > > > switches. The other part of the USB implementation exposes UCSI.
> > > > 
> > > > The altmode implementation provides two things:
> > > > - A drm_bridge, per connector, which can be tied (of_graph) to a
> > > >     DisplayPort instance, and will invoke HPD notifications on the
> > > >     drm_bridge, based on notification messages thereof.
> > > > 
> > > > - Acquire typec_switch and typec_mux handles through the of_graph and
> > > >     signal the remotes when notifications of state changes occur. Linking
> > > >     this to the FSA4480, is sufficient to get USB/DP combo (2+2 lanes)
> > > >     working on e.g. SM8350 HDK.
> > > >     Work in progress patches also exists for teaching QMP about
> > > >     orientation switching of the SS lines, but it seems this needs to be
> > > >     rebased onto the refactored QMP driver.
> > > >     I also have patches for QMP to make it switch USB/DP combo -> 4-lane
> > > >     DP, which allow 4k support without DSC, unfortunately switch back to
> > > >     USB has not been fully reliable, so this requires some more work
> > > >     (downstream involves DWC3 here as well, to reprogram the PHY).
> > > 
> > > Oki doki that makes sense and is pretty much in-line with what I thought.
> > > 
> > > We still have a bunch of typec-mux and phy work to do even with adsp/glink
> > > doing the TCPM.
> > > 
> > 
> > Correct, the registration of QMP as a typec_switch and typec_mux and
> > handling of respective notification remains open and should (by design)
> > be independent of the TCPM implementation.
> > 
> > In particular the orientation switching is an itch worth scratching at
> > this time. But when the DPU becomes capable of producing 4k@60 output it
> > would obviously be nice to have the whole shebang :)
> 
> Did you try it with the wide planes patchset at [1]? I was able to get
> stable 4k@30 on RB3 (being limited only by the DSI-HDMI bridge).
> 
> [1] https://lore.kernel.org/linux-arm-msm/20221229191856.3508092-1-dmitry.baryshkov@xxxxxxxxxx/
> 

I have not done so, as the patches I had for switching to 4-lane DP
output needs to be rewritten since the refactoring.

I had no problem doing 4k@30 prior to your efforts, so I'm interested in
validating the changes you've made.

Thanks,
Bjorn

> -- 
> With best wishes
> Dmitry
>