Re: Using TCPM for ports without Power Delivery support

[Guenter Roeck <linux@roeck-us.net>]

On 2/17/21 7:28 AM, Cristian.Birsan@microchip.com wrote:
> Hi Heikki,
> 
> On 2/16/21 11:12 AM, Heikki Krogerus wrote:
>>
>> Hi Cristian,
>>
>> On Mon, Feb 15, 2021 at 05:25:29PM +0000, Cristian.Birsan@microchip.com wrote:
>>> My name is Cristian and I'm working on bringing up a USB Type-C Port Controller
>>> (TCPC) without Power Delivery support which is intended to work with USB 2.0
>>> Host/Device.
>>>
>>> The IP is integrated into one of Microchip's SoCs, it is memory-mapped and it
>>> was designed based on USB Type-C Cable and Connector specification revision 1.2.
>>>
>>> In brief, it has support for detecting the threshold voltages on CC1, CC2 lines,
>>> control of the current source (Ip), and pull-down resistors (Rd). The management
>>> of the controller is to be implemented in software (it is not autonomous).
>>>
>>> Having in mind that the controller uses proprietary registers, I chose to
>>> implement it using TCPM directly and skip the TCPC Interface.
>>>
>>> For the beginning, I would like to enable simple use cases like the ones
>>> described in Connection State Diagram: Source and Connection State Diagram: Sink
>>> from USB Type-C Cable and Connector Specification.
>>>
>>> Some of the problems that I encountered until now are:
>>>
>>> 1. tcpm_register_port() fails if set_pd_rx(), pd_transmit() or set_vconn()
>>> functions are missing.
>>>
>>> 2. the port capabilities are specified in the connector DT bindings only through
>>> PDOs, even though PDOs are specific to PD mode.
>>>
>>> 3. once I was able to start the TCPM state machine, it called pd_transmit() in
>>> the process to negotiate the capabilities. For my case I used a dummy function
>>> just to be able to register the port.
>>>
>>> Please let me know what you think and if you have any advice. Am I going in the
>>> right direction or is there a better way to implement this?
>>
>> Don't bother with tcpm if you don't have PD support. Just register
>> your port(s) and the partners directly with the connector class:
>> https://www.kernel.org/doc/html/latest/driver-api/usb/typec.html
>>
>> You can use the driver for the TI HD3SS3220 controller as an example
>> how to do that (drivers/usb/typec/hd3ss3220.c). That thing is also
>> just USB Type-C PHY without PD support just like your port controller.
>>
> 
> Thank you for the suggestion. I had a look at the driver you mentioned and also
> other drivers from the same folder. The chips have logic embedded into
> them to handle CC lines and VBUS while the controller on which I'm working
> provides basic access to CC lines and it is up to the software to drive it.
> VBUS is to be detected/enabled through a standard GPIO.
> 
> The reason for which I tried to use TCPM is that I need to implement in software
> the Sink/Source Connection State Diagrams, CC debounce, and VBUS management.
> I tried to avoid code duplication but on the other hand, my use case does not
> involve PD.
> 
> If there are better chances to upstream the driver using just the connector
> class, I'll move forward with this direction. I just wanted to make sure I
> explained correctly may use case before implementing it.
> 

Personally I think it would be a bad idea to make the mandatory functions optional.
PD is an integral part of tcpm, after all. How about implementing dummy functions
as needed ?

Thanks,
Guenter