Re: [PATCH] usb: dwc3: gadget: lower informal user notifaction dequeue operation

[Marco Felsch]

Hi,

On 23-03-23, Thinh Nguyen wrote:
> Hi,
> 
> On Thu, Mar 23, 2023, Marco Felsch wrote:
> > Printing an error message during usb_ep_dequeue() is more confusing than
> > helpful since the usb_ep_dequeue() could be call during unbind() just
> > in case that everything is canceld before unbinding the driver. Lower
> > the dev_err() message to dev_dbg() to keep the message for developers.
> > 
> > Fixes: fcd2def66392 ("usb: dwc3: gadget: Refactor dwc3_gadget_ep_dequeue")
> > Signed-off-by: Marco Felsch <m.felsch@xxxxxxxxxxxxxx>
> > ---
> >  drivers/usb/dwc3/gadget.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
> > index 89dcfac01235f..6699db26cc7b5 100644
> > --- a/drivers/usb/dwc3/gadget.c
> > +++ b/drivers/usb/dwc3/gadget.c
> > @@ -2106,7 +2106,7 @@ static int dwc3_gadget_ep_dequeue(struct usb_ep *ep,
> >  		}
> >  	}
> >  
> > -	dev_err(dwc->dev, "request %pK was not queued to %s\n",
> > +	dev_dbg(dwc->dev, "request %pK was not queued to %s\n",
> >  		request, ep->name);
> >  	ret = -EINVAL;
> >  out:
> > -- 
> > 2.30.2
> > 
> 
> How were you able to reproduce this error message?

We use the driver within barebox where we do have support for fastboot.
During the driver unbind usb_ep_dequeue() is called which throw this
error.

> During unbind(), the function driver would typically call to
> usb_ep_disable(). Before the call usb_ep_disable() completes, all queued
> and incompleted requests are expected to be returned with -ESHUTDOWN.

So the unbind() function driver should use usb_ep_disable() instead of
usb_ep_dequeue()?

> For you to see this error, this means that the function driver issued
> usb_ep_dequeue() to an already disabled endpoint, and the request was
> probably already given back.

The unbind() just calls usb_ep_dequeue() which isn't forbidden according
the API doc. We just want to ensure that the request is cancled if any.

> Even though this error message is not critical and shouldn't affect the
> driver's behavior, it's better to fix the function driver to handle this
> race.

As you have pointed out: 'it is not criticial' and therefore we shouldn't
use dev_err() for non crictical information since this can cause
user-space confusion.

Regards,
  Marco

> 
> Thanks,
> Thinh