This patchset will revert a recent fix patch which will introduce another
regression, and fix the true underlying issue.

Arnaud pushed a fix patch commit 1366cd228b0c ("tcpm: allow looking for
role_sw device in the main node") because commit 2d8713f807a4
("tcpm: switch check for role_sw device with fw_node") swap the order of
usb_role_switch_get() and fwnode_usb_role_switch_get() and this cause the
fwnode_usb_role_switch_get() return "-EPROBE_DEFER" all the time in his
usecase.

The said dts use below property to get role switch device: 

  fusb0: typec-portc@22 {
  	  compatible = "fcs,fusb302";
  	  usb-role-switch = <&typec_extcon_bridge>;
  	  ...
   
  	  connector {
  	  }
  }

So the fix patch works because usb_role_switch_get() will correctly find
the role switch device in case fwnode_usb_role_switch_get() is failed.
However, usb_role_switch_get() shouldn't be called at all in this case
because tcpc->fwnode is already correctly set to @connector. And worse
more, this bring another issue that is once fwnode_usb_role_switch_get()
return "-EPROBE_DEFER" this info may be lost if usb_role_switch_get()
then return NULL.

After further research into this issue, finally I found
usb_role_switch_is_parent() not work properly when it parse non-boolean
"usb-role-switch" property which is for back compatibility purpose. To fix
the issue, patch#2 is added.

Xu Yang (2):
  Revert "tcpm: allow looking for role_sw device in the main node"
  usb: roles: get usb role switch from parent only for usb-b-connector

 drivers/usb/roles/class.c     | 7 ++++++-
 drivers/usb/typec/tcpm/tcpm.c | 2 +-
 2 files changed, 7 insertions(+), 2 deletions(-)

-- 
2.34.1

Mon, Mar 09, 2026 at 03:43:11PM +0800, Xu Yang kirjoitti:
> This patchset will revert a recent fix patch which will introduce another
> regression, and fix the true underlying issue.
> 
> Arnaud pushed a fix patch commit 1366cd228b0c ("tcpm: allow looking for
> role_sw device in the main node") because commit 2d8713f807a4
> ("tcpm: switch check for role_sw device with fw_node") swap the order of
> usb_role_switch_get() and fwnode_usb_role_switch_get() and this cause the
> fwnode_usb_role_switch_get() return "-EPROBE_DEFER" all the time in his
> usecase.
> 
> The said dts use below property to get role switch device: 
> 
>   fusb0: typec-portc@22 {
>   	  compatible = "fcs,fusb302";
>   	  usb-role-switch = <&typec_extcon_bridge>;
>   	  ...
>    
>   	  connector {
>   	  }
>   }
> 
> So the fix patch works because usb_role_switch_get() will correctly find
> the role switch device in case fwnode_usb_role_switch_get() is failed.
> However, usb_role_switch_get() shouldn't be called at all in this case
> because tcpc->fwnode is already correctly set to @connector. And worse
> more, this bring another issue that is once fwnode_usb_role_switch_get()
> return "-EPROBE_DEFER" this info may be lost if usb_role_switch_get()
> then return NULL.
> 
> After further research into this issue, finally I found
> usb_role_switch_is_parent() not work properly when it parse non-boolean
> "usb-role-switch" property which is for back compatibility purpose. To fix
> the issue, patch#2 is added.
> 
> Xu Yang (2):
>   Revert "tcpm: allow looking for role_sw device in the main node"
>   usb: roles: get usb role switch from parent only for usb-b-connector
> 
>  drivers/usb/roles/class.c     | 7 ++++++-
>  drivers/usb/typec/tcpm/tcpm.c | 2 +-
>  2 files changed, 7 insertions(+), 2 deletions(-)

For the series:

Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>

Thanks Xu!

-- 
heikki