On Fri, Aug 17, 2012 at 11:10:19AM +0200, Roland Stigge wrote:
> Hi,
>
> On 08/16/2012 06:16 PM, Roland Stigge wrote:
> > On 08/16/2012 06:05 PM, Sebastian Andrzej Siewior wrote:
> >>> --- linux-2.6.orig/drivers/usb/gadget/lpc32xx_udc.c
> >>> +++ linux-2.6/drivers/usb/gadget/lpc32xx_udc.c
> >>> @@ -2987,14 +2986,14 @@ static irqreturn_t lpc32xx_usb_vbus_irq(
> >>> return IRQ_HANDLED;
> >>> }
> >>>
> >>> -static int lpc32xx_start(struct usb_gadget_driver *driver,
> >>> - int (*bind)(struct usb_gadget *))
> >>> +static int lpc32xx_start(struct usb_gadget *gadget,
> >>> + struct usb_gadget_driver *driver)
> >>> {
> >>> - struct lpc32xx_udc *udc =&controller;
> >>
> >> I assume controller is a global var created at probe time and could be
> >> removed now, right?
> >
> > Yes!
>
> Well ;-) looking more closely into it, I'd like to keep it for now: It
> is a more complex statically pre-initialized struct, finally being used
> in probe() for more dynamic initialization, and it ends up being used by
> many other functions, including lpc32xx_start() where accessing it now
> via container_of() is just a bit more elegant than a direct &controller
> of the global variable.
>
> Also, since this device is a single controller in the LPC32xx SoC, I
> would keep it until some other silicon uses several of this IP core
> (which I doubt), at which point we would probably still keep the (global
> static) controller and memcpy it to a dynamically allocated struct.
>
> Sounds reasonable?
no it doesn't. Please remove that static global. Sorry but one of the
goals with udc_start/udc_stop was really to get rid of all those
nonsensical static globals.
--
balbi
Attachment:
signature.asc
Description: Digital signature
