Re: [PATCH v4 08/13] phy: qcom: qmp-usbc: Add USB/DP switchable PHY clk register

[Xiangxu Yin <xiangxu.yin@oss.qualcomm.com>]

On 9/15/2025 6:14 PM, Dmitry Baryshkov wrote:
> On Mon, Sep 15, 2025 at 06:02:19PM +0800, Xiangxu Yin wrote:
>> On 9/12/2025 8:08 PM, Dmitry Baryshkov wrote:
>>> On Fri, Sep 12, 2025 at 08:00:14PM +0800, Xiangxu Yin wrote:
>>>> On 9/12/2025 6:19 PM, Dmitry Baryshkov wrote:
>>>>> On Thu, Sep 11, 2025 at 10:55:05PM +0800, Xiangxu Yin wrote:
>>>>>> Add USB/DP switchable PHY clock registration and DT parsing for DP offsets.
>>>>>> Extend qmp_usbc_register_clocks and clock provider logic to support both
>>>>>> USB and DP instances.
>>>>>>
>>>>>> Signed-off-by: Xiangxu Yin <xiangxu.yin@oss.qualcomm.com>
>>>>>> ---
>>>>>>  drivers/phy/qualcomm/phy-qcom-qmp-usbc.c | 208 +++++++++++++++++++++++++++++--
>>>>>>  1 file changed, 195 insertions(+), 13 deletions(-)
>>>>>> +	default:
>>>>>> +		return 0;
>>>>>> +	}
>>>>>> +}
>>>>>> +
>>>>>> +static int qmp_usbc_register_clocks(struct qmp_usbc *qmp, struct device_node *np)
>>>>>> +{
>>>>>> +	int ret;
>>>>>>  
>>>>>> -	ret = of_clk_add_hw_provider(np, of_clk_hw_simple_get, &fixed->hw);
>>>>>> +	ret = phy_pipe_clk_register(qmp, np);
>>>>>>  	if (ret)
>>>>>>  		return ret;
>>>>>>  
>>>>>> -	/*
>>>>>> -	 * Roll a devm action because the clock provider is the child node, but
>>>>>> -	 * the child node is not actually a device.
>>>>>> -	 */
>>>>>> -	return devm_add_action_or_reset(qmp->dev, phy_clk_release_provider, np);
>>>>>> +	if (qmp->dp_serdes != 0) {
>>>>>> +		ret = phy_dp_clks_register(qmp, np);
>>>>>> +		if (ret)
>>>>>> +			return ret;
>>>>>> +	}
>>>>>> +
>>>>>> +	return devm_of_clk_add_hw_provider(qmp->dev, qmp_usbc_clks_hw_get, qmp);
>>>>> Do you understand what did the comment (that you've removed) say? And
>>>>> why?
>>> And this was ignored :-(
>> Sorry for missing this part.
>>
>> For USB-C PHY, the legacy implementation only supports USB with a single
>> device node. The new driver for USB and DP also uses a single device node.
> There is no 'new driver'. It's about DT.
>
>> The function devm_of_clk_add_hw_provider internally handles both
>> of_clk_add_hw_provider and devres_add, and supports automatic resource
>> release.
>>
>> So I think using devm_of_clk_add_hw_provider allows us to remove
>> of_clk_add_hw_provider and devm_add_action_or_reset.
> Which node is passed to of_clk_add_hw_provider() in the legacy DT case?
> Which node is passed to of_clk_add_hw_provider() by
> devm_of_clk_add_hw_provider()?


Ohh, legacy is child node and devm is dev->of_node.

Will add that back for compatibility.


>> For combo PHY, the legacy implementation uses two device nodes: dp_np and
>> usb_np. To maintain forward compatibility, we need to keep support for
>> both nodes and retain the related logic.