Re: [PATCH v4 0/3] Skip phy initialization for DWC3 USB Controllers

[Pavan Kondeti <quic_pkondeti@quicinc.com>]

Hi Thinh,

On Tue, Apr 26, 2022 at 01:12:17AM +0000, Thinh Nguyen wrote:
> Hi,
> 
> Pavan Kondeti wrote:
> > Hi Mathias,
> > 
> > On Thu, Apr 21, 2022 at 10:30:52AM +0530, Sandeep Maheswaram wrote:
> >> Runtime suspend of phy drivers was failing from DWC3 driver as
> >> runtime usage value is 2 because the phy is initialized from
> >> DWC3 core and HCD core.
> >> Some controllers like DWC3 and CDNS3 manage phy in their core drivers.
> >> This property can be set to avoid phy initialization in HCD core.
> >>
> >> v4:
> >> Added the device tree binding patch in the series.
> >>
> >> v3:
> >> Coming back to this series based on discussion at below thread
> >> https://urldefense.com/v3/__https://patchwork.kernel.org/project/linux-arm-msm/patch/1648103831-12347-4-git-send-email-quic_c_sanm@quicinc.com/__;!!A4F2R9G_pg!fykTNTBuKk9ci6zKdcuQNbuZQdVi_HekU3jetzud-PQVhbRaVhhZHKz0k_LfG0cgwaX4bQM5bLI0ep6tYyikgvYK7b5SdA$ 
> >> Dropped the dt bindings PATCH 1/3 in v2
> >> https://urldefense.com/v3/__https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/__;!!A4F2R9G_pg!fykTNTBuKk9ci6zKdcuQNbuZQdVi_HekU3jetzud-PQVhbRaVhhZHKz0k_LfG0cgwaX4bQM5bLI0ep6tYyikgva2VXahOQ$  
> >>
> >> v2:
> >> Updated the commit descriptions.
> >> Changed subject prefix from dwc to dwc3.
> >> Increased props array size.
> >>
> >>
> >> Sandeep Maheswaram (3):
> >>   dt-bindings: usb: usb-xhci: Add bindings for usb-skip-phy-init
> >>     property
> >>   usb: host: xhci-plat: Add device property to set XHCI_SKIP_PHY_INIT
> >>     quirk
> >>   usb: dwc3: host: Set the property usb-skip-phy-init
> >>
> >>  Documentation/devicetree/bindings/usb/usb-xhci.yaml | 4 ++++
> >>  drivers/usb/dwc3/host.c                             | 4 +++-
> >>  drivers/usb/host/xhci-plat.c                        | 3 +++
> >>  3 files changed, 10 insertions(+), 1 deletion(-)
> >>
> > 
> > This is the latest series with bindings added as per Greg's comment. Can you
> > please pick up this series if you don't have any further comments.
> > 
> 
> We've had this conversation going on for a while. Seems there's no good
> one solution with everyone fully getting on-board.
> 
> I've tried to get some of the quirks out before also, but ran into the
> same problem. [1]
> 
> As Mathias noted [2] before, maybe we can create a new xhci-snps
> platform glue driver.
> 
> The problem with the current implementation is passing dwc3's related
> info to xhci-plat generic driver is very clunky. We can teach the new
> glue driver with all the info necessary to drive the controller.
> 
> We can just pass the controller's version (and subversion) as a property
> for platform device. This way, we can:
> 
> 1) Separate the quirks from xhci-plat glue. Most common quirks can be
> detected just base on the controller's version
> 
> 2) Avoid having to create duplicate "snps,*" properties
> 
> 3) Get access to the common xhci quirk flags while maintain abstraction
> 
> 4) Potentially add compatibility string as part of the controller's
> version and let the glue driver handle the rest
> 
> 5) Reduce introducing new "quirks" in the future
> 
> I can get started with this. Let me know if you have any comment.

Sorry, could not reply earlier. The proposal sounds good to me.

The xhci-plat is a thin wrapper, so having a separate wrapper for SNPS
controller is definitely not an overkill and gives lot of flexibility
in abstracting dwc3 specifics. Also dwc3/host.c becomes just a platform
device creation wrapper and xHC specifics are completely taken out.

Thanks,
Pavan