Re: [PATCH] Revert "tcpm: allow looking for role_sw device in the main node"

[Arnaud Ferraris <arnaud.ferraris@collabora.com>]

Hi Xu,

Le 25/02/2026 à 03:57, Xu Yang a écrit :
> Hi Arnaud,
> 
> On Tue, Feb 24, 2026 at 12:33:33PM +0100, Arnaud Ferraris wrote:
>> Hi,
>>
>> Le 24/02/2026 à 12:01, Xu Yang a écrit :
>>> This reverts commit 1366cd228b0c67b60a2c0c26ef37fe9f7cfedb7f.
>>
>> I believe a plain revert isn't the right solution here, as we'll get to the
>> same point as we were before 1366cd228b0c, where some devices stopped
>> working properly with newer kernels.
> 
> I don't think 1366cd228b0c fix the real root problem because the description
> should be wrong in the commit message. If -EPROBE_DEFER is returned by
> fwnode_usb_role_switch_get(), the ports node should be in connector node
> instead of tcpc node. However, you get the error when ports in tcpc node.
> 
> Could you double check the issue, so we can find a proper solution and avoid
> the further regression?

Sure, I'll come up with more details asap, either tomorrow or early next 
week.

Best regards,
Arnaud

> 
>>
>>>
>>> The fwnode_usb_role_switch_get() returns NULL only if no connection is
>>> found, returns ERR_PTR(-EPROBE_DEFER) if connection is found but deferred
>>> probe is needed, or a valid pointer of usb_role_switch.
>>>
>>> When switching from NULL check to IS_ERR_OR_NULL(), usb_role_switch_get()
>>> will return NULL pointer which will override ERR_PTR(-EPROBE_DEFER) which
>>> is returned by fwnode_usb_role_switch_get(). Then usb role switch can't be
>>> obtained because the defer probe info is lost. So the unique error should
>>> not be regarded the same as NULL.
>>>
>>> Fixes: 1366cd228b0c ("tcpm: allow looking for role_sw device in the main node")
>>> Cc: stable@vger.kernel.org
>>> Signed-off-by: Xu Yang <xu.yang_2@nxp.com>
>>>
>>> ---
>>> Also correct a description in 1366cd228b0c ("tcpm: allow looking for
>>> role_sw device in the main node"), if the ports are defined in the tcpc
>>> main node, NULL pointer is returned by fwnode_usb_role_switch_get()
>>> instead of an error.
>>> ---
>>>    drivers/usb/typec/tcpm/tcpm.c | 2 +-
>>>    1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/usb/typec/tcpm/tcpm.c b/drivers/usb/typec/tcpm/tcpm.c
>>> index 1d2f3af034c5..8e0e14a2704e 100644
>>> --- a/drivers/usb/typec/tcpm/tcpm.c
>>> +++ b/drivers/usb/typec/tcpm/tcpm.c
>>> @@ -7890,7 +7890,7 @@ struct tcpm_port *tcpm_register_port(struct device *dev, struct tcpc_dev *tcpc)
>>>    	port->partner_desc.identity = &port->partner_ident;
>>>    	port->role_sw = fwnode_usb_role_switch_get(tcpc->fwnode);
>>> -	if (IS_ERR_OR_NULL(port->role_sw))
>>> +	if (!port->role_sw)
>>
>> It might be worth saving the error and restoring it after the call to
>> usb_role_switch_get() instead, something like:
>>
>> 	if (IS_ERR_OR_NULL(port->role_sw)) {
>> 		err = PTR_ERR(port->role_sw);
>> 		port->role_sw = usb_role_switch_get(port->dev);
>> 		if (!port->role_sw)
>> 			port->role_sw = ERR_PTR(err);
>> 	}
> 
> It works but we'd better to get the thing clear firstly.
> 
> Thanks,
> Xu Yang