On Mon, Jan 19, 2015 at 11:29:22AM -0500, Alan Stern wrote: > On Mon, 19 Jan 2015, Peter Chen wrote: > > > > > An idea just struck me... Instead of looping through all the udc's to > > > > find the right one, why not simply store a pointer to the udc in struct > > > > usb_gadget? > > > > > > > > We still have code to find usb_gadget_driver to iterate udc_list, it > > is better to change all, we need to have solution to consolidate > > struct usb_udc, struct usb_gadget, and struct usb_gadget_drver. > > I have no good solution now. > > I think it's okay to iterate through the list to find a > usb_gadget_driver. But it's silly to iterate to find the usb_udc > corresponding to a usb_gadget. Felipe, what's your opinion? > > > > > Also, it looks like there's a bug in usb_add_gadget_udc_release() in > > > > udc-core.c. The error pathway (between err3 and err2) does not undo > > > > the > > > > > > > > ret = device_register(&gadget->dev); > > > > > > > > call. There's a put_device() call but no device_del(). > > > > > > good point :-) Do you want to send a patch for that ? > > > > > > > It looks like there is another issue in error pathway, the > > put_device(&udc->dev) at err3 should be moved to err4 since > > the get_device is called at device_add. > > No, the get_device is done by device_initialize, so the code is correct > as it stands. See the kerneldoc for device_initialize() in > drivers/base/core.c. > > Alan Stern > Yes, you are right, just device_initialize calls kobject_init to set refcount as 1, not call get_device directly. -- Best Regards, Peter Chen -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
