[v2,0/2] usb: dwc2: Use clk bulk API for supporting multiple clocks

Message

Sascha Hauer Jan. 25, 2021, 9:38 a.m. UTC

Currently the dwc2 driver only supports a single clock. I have a board
here which has a dwc2 controller with a somewhat special clock routing
to the phy. Both the dwc2 controller and the ULPI phy get their phy
clock from a SI5351 clock generator. This clock generator has multiple
clock outputs which each is modelled as a separate clk in Linux.
Unfortunately the clock to the phy and the clock to the dwc2 core are on
two different output pins of the SI5351, so we have two clocks which
must be enabled.  The phy is driven by the usb-nop-xceiver driver which
supports a single clock. My first approach was to add support for a
second clock to that driver, but technically the other clock is
connected to the dwc2 core, so instead I added support for a second
clock to the dwc2 driver.  This can easily be archieved with the clk
bulk API as done in this series.

Changes since v1:
- Add minItems to dec2 clock/clock-names property to avoid
  dt_binding_check warning

Sascha Hauer (2):
  dt-bindings: usb: dwc2: Add support for additional clock
  usb: dwc2: use clk bulk API for supporting additional clocks

 .../devicetree/bindings/usb/dwc2.yaml          |  5 ++++-
 drivers/usb/dwc2/core.h                        |  2 ++
 drivers/usb/dwc2/platform.c                    | 18 ++++++++----------
 3 files changed, 14 insertions(+), 11 deletions(-)

Comments

Rob Herring (Arm) Feb. 9, 2021, 4:54 p.m. UTC | #1

On Mon, Jan 25, 2021 at 10:38:23AM +0100, Sascha Hauer wrote:
> Currently the dwc2 driver only supports a single clock. I have a board
> here which has a dwc2 controller with a somewhat special clock routing
> to the phy. Both the dwc2 controller and the ULPI phy get their phy
> clock from a SI5351 clock generator. This clock generator has multiple
> clock outputs which each is modelled as a separate clk in Linux.
> Unfortunately the clock to the phy and the clock to the dwc2 core are on
> two different output pins of the SI5351, so we have two clocks which
> must be enabled.  The phy is driven by the usb-nop-xceiver driver which
> supports a single clock. My first approach was to add support for a
> second clock to that driver, but technically the other clock is
> connected to the dwc2 core, so instead I added support for a second
> clock to the dwc2 driver.  This can easily be archieved with the clk
> bulk API as done in this series.

Where is the usb-nop-xceiver single clock coming from? 

Maybe you shouldn't be using usb-nop-xceiver? There is a ULPI binding 
though that alone doesn't really help you.

Rob

Sascha Hauer Feb. 10, 2021, 8:42 a.m. UTC | #2

On Tue, Feb 09, 2021 at 10:54:24AM -0600, Rob Herring wrote:
> On Mon, Jan 25, 2021 at 10:38:23AM +0100, Sascha Hauer wrote:
> > Currently the dwc2 driver only supports a single clock. I have a board
> > here which has a dwc2 controller with a somewhat special clock routing
> > to the phy. Both the dwc2 controller and the ULPI phy get their phy
> > clock from a SI5351 clock generator. This clock generator has multiple
> > clock outputs which each is modelled as a separate clk in Linux.
> > Unfortunately the clock to the phy and the clock to the dwc2 core are on
> > two different output pins of the SI5351, so we have two clocks which
> > must be enabled.  The phy is driven by the usb-nop-xceiver driver which
> > supports a single clock. My first approach was to add support for a
> > second clock to that driver, but technically the other clock is
> > connected to the dwc2 core, so instead I added support for a second
> > clock to the dwc2 driver.  This can easily be archieved with the clk
> > bulk API as done in this series.
> 
> Where is the usb-nop-xceiver single clock coming from?

It's coming from the SI5351 clock generator

> 
> Maybe you shouldn't be using usb-nop-xceiver? There is a ULPI binding 
> though that alone doesn't really help you.

The clock that feeds the phy is handled by the usb-nop-xceiver just
fine, it's the clock that is fed to the DWC2 controller that I need
support for. I hope my other mail makes this clearer a bit.

Sascha