> -----Original Message----- > From: Marek Vasut <ma...@denx.de> > Sent: Saturday, July 20, 2024 22:12 > > Can you please send an RFC patch ? > > > I see recently there are many DM related reworking, please forgive me > > if it's already on the Todo list. > > It is not on the ToDo list, no. > > +CC Mattijs
Hi, this issue only affects usb_ether, as g_dnl will enter event loop right after driver register so the host won't see an un responsive device. I think one solution is revert half of 718f1d41 to move usb_gadget_register_driver()/usb_gadget_unregister_driver() back to usb_eth_start()/usb_eth_stop(). It doesn't affect the ability of bind/unbind mentioned by the commit's description. In fact by reverting it usb_ether will behaves more like running 'ums' or 'dfu', since running a ethernet command like 'dhcp' will do register->event loop->unregister just like 'ums' does.
