[PATCH v2] usb: cdns3: use VBUS Valid to determine role for dr_mode OTG

[Marek Vasut]

On 2/14/26 5:36 AM, Siddharth Vadapalli wrote:
> Hello Marek,
> 
> On Fri, 2026-02-13 at 16:04 +0100, Marek Vasut wrote:
>> On 2/13/26 6:15 AM, Siddharth Vadapalli wrote:
>>
>> [...]
>>
>>> +++ b/drivers/usb/cdns3/core.c
>>> @@ -392,6 +392,63 @@ static const struct udevice_id cdns3_ids[] = {
>>>    	{ },
>>>    };
>>>    
>>> +/*
>>> + * The VBUS Valid Bit in the OTG Status register can be used to determine
>>> + * the role. When VBUS Valid is set, it indicates that a USB Host is supplying
>>> + * power, so the Controller should assume the PERIPHERAL role. If it isn't set,
>>> + * it indicates the absence of a USB Host, so the Controller should assume the
>>> + * HOST role. If the OTG Status register is inaccessible, return an error.
>>> + */
>>> +static int cdns3_get_otg_mode(ofnode node
>>
>> Pass struct udevice *parent instead of node here ...
>>
>>> , enum usb_dr_mode *mode)
>>> +{
>>> +	struct cdns3 cdns, *cdnsp;
>>> +	void __iomem *otg_regs;
>>> +	fdt_addr_t otg_addr;
>>> +	int otg_reg_index;
>>> +	int vbus;
>>> +
>>> +	otg_reg_index = ofnode_stringlist_search(node, "reg-names", "otg");
>>> +	if (otg_reg_index < 0) {
>>> +		printf("%s: dr_mode is otg but otg register name is missing\n", __func__);
>>> +		return -ENOENT;
>>> +	}
>>> +
>>> +	otg_addr = ofnode_get_addr_index(node, otg_reg_index);
>>> +	if (otg_addr == FDT_ADDR_T_NONE) {
>>> +		printf("%s: address of otg registers is unspecified\n", __func__);
>>> +		return -ENOENT;
>>> +	}
>>> +
>>> +	otg_regs = map_physmem(otg_addr, 0, MAP_NOCACHE);
>>> +	if (!otg_regs) {
>>> +		printf("%s: failed to map otg registers\n", __func__);
>>> +		return -EINVAL;
>>> +	}
>>
>> ... and replace all that ^ with devfdt_remap_addr_name(parent, "otg") .
>> That should work.
> 
> It doesn't work because the 'parent' node doesn't have the 'otg' registers.
> It is the child node which has it. The device-tree nodes for parent and
> child are:
> 
> 	usbss0: usb at 4104000 {
> 							=> PARENT
> 		bootph-all;
> 		compatible = "ti,j721e-usb";
> 		reg = <0x00 0x4104000 0x00 0x100>;
> 		dma-coherent;
> 
> 		[TRIMMED]
> 
> 		usb0: usb at 6000000
> {							=> CHILD
> 			bootph-all;
> 			compatible = "cdns,usb3";
> 			reg = <0x00 0x6000000 0x00 0x10000>,
> 			      <0x00 0x6010000 0x00 0x10000>,
> 			      <0x00 0x6020000 0x00 0x10000>;
> 			reg-names = "otg", "xhci",
> "dev";			=> OTG Register is defined here
> 
> 			[TRIMMED]
> 	};
> 
> The 'udevice' for the child node is created in the cdns3_bind() function
> __after__ the function in this patch is invoked. Therefore, we don't yet
> have the child device which is created by:
> 
> 		device_bind_driver_to_node(parent, driver, name, node,
> &dev);
> 
> I did consider the possibility of using the standard APIs
> (devfdt_remap_addr_name), but since the child device doesn't exist yet,
> they cannot be used. Hence, I had to implement custom logic.
Maybe it would be possible to create a temporary udevice on stack and 
pass it to devfdt_remap_addr_name() ? I think this should be complete 
enough, although I might have missed something:

"
struct udevice cdev = { .parent = parent; };
dev_set_ofnode(cdev, node);
"

devfdt_remap_addr_name() internally calls dev_of_offset(dev) on the 
cdev, tests is device_get_uclass_id(dev->parent) == UCLASS_SIMPLE_BUS 
(it is not) and that is it.