[U-Boot] [PATCH 3/4] usb: dwc3: add support for 16 bit UTMI+ interface

[Marek Vasut]

On 08/18/2016 11:04 AM, Kever Yang wrote:
> Hi Marek,

Hi,

> On 08/17/2016 04:07 PM, Marek Vasut wrote:
>> On 08/17/2016 04:06 AM, Kever Yang wrote:
>>> Hi Marek,
>>>
>>> On 08/17/2016 09:39 AM, Marek Vasut wrote:
>>>> On 08/17/2016 03:31 AM, Kever Yang wrote:
>>>>> Hi Marek,
>>>> Hi,
>>>>
>>>>> On 08/16/2016 09:18 PM, Marek Vasut wrote:
>>>>>> On 08/16/2016 12:03 PM, Kever Yang wrote:
>>>>>>> The dwc3 controller is using 8 bit UTMI+ interface for USB2.0 PHY,
>>>>>>> add one MACRO CONFIG_USB_DWC3_USB2PHY_16BIT to support 16 bit
>>>>>>> UTMI+ interface on some SoCs like Rockchip rk3399.
>>>>>>>
>>>>>>> Signed-off-by: Kever Yang <kever.yang at rock-chips.com>
>>>>>> This should be configured from either DT or Kconfig option.
>>>>> This setting in kernel is from DT, but we do not use DT for gadget
>>>>> mode
>>>>> in U-Boot.
>>>>> How about we add a quirk for it like u2phy_utmi_width_quirk in both
>>>>> dwc3_device
>>>>> and dwc3 struct for device mode?
>>>>>
>>>>> For Host mode, we can get from DT and also need a quirk to store it.
>>>> Surely, if you can get this info from DT for host mode, you can also
>>>> get
>>>> it from DT for gadget mode, yes ?
>>> No, for gadget mode, the controller is init by board_usb_init() called
>>> from cmd like fastboot,
>>> there is no uclass driver for a udc driver and no one parse the DT
>>> currently.
>> Which doesn't imply you cannot parse the DT though, correct ?
>> I would _much_ rather see you parsing DT over adding new ad-hoc compile
>> time config option.
> 
> Sorry, I don't understand what do you mean by "adding new ad-hoc compile
> time config option".

I mean adding CONFIG_USB_DWC3_USB2PHY_16BIT instead of parsing DT .

> Do you mean (a) add fdt_get_property() to parse DT only for this setting?
> Or (b) implement a new driver for gadget mode with driver-model support
> which
> bind the device during DT instead of board_usb_init()?

I think the former would be easier, yes ?

> The case (a) would be easy while case (b) suppose to be a big update.
> The source of driver for dwc3 gadget is in drivers/usb/dwc3 and the
> source of driver for
> dwc3 host is in drivers/usb/host, we will have to merge this two driver
> into one in case (b)
> bacause they are using the some controller which means the same dts node.

I believe drivers/usb/dwc3 is pulled from Linux, so this is the only
codebase which we should ever use to add host mode to.

> I would like to see these two driver merge into one like what kernel do,
> but I think it need
> someone who very familiar with both dwc3 host and gadget to port the
> driver from kernel.

Implementing (a) is fine, while (b) would be excellent.

> Thanks,
> - Kever
>>
>>> You can see my patch "[PATCH 2/4] board: evb-rk3399: add api to support
>>> dwc3 gadget"
>>> for how it initialized.
>>>
>>> Thanks,
>>> - Kever
>> [...]
>>
> 
> 


-- 
Best regards,
Marek Vasut