Re: [PATCH v4] USB: Fix device driver race

[Bastien Nocera <hadess@hadess.net>]

On Fri, 2020-07-24 at 13:08 -0400, Alan Stern wrote:
> On Fri, Jul 24, 2020 at 06:52:31PM +0200, Bastien Nocera wrote:
> > On Fri, 2020-07-24 at 11:27 -0400, Alan Stern wrote:
> > > > > +               if (new_udriver->match || new_udriver-
> > > > > >id_table)
> > > > 
> > > > But match check is incorporated in the loop function.
> > > 
> > > Agreed, this test is redundant.  However, we should test that 
> > > new_udriver != &usb_generic_driver.
> > 
> > Do you really want to loop over every USB device when you know for
> > a
> > fact that not a single one will match?
> 
> Think of it the other way around: How often will anybody load a 
> specialized USB device driver that doesn't have a match function or
> ID 
> table?  It wouldn't match any devices!
> 
> > I guess it's unlikely, the generic driver would be loaded before
> > any
> > device,
> 
> Since it's built into usbcore, I guess that's true.
> 
> >  and the specialised drivers need to be able to selected, so
> > I've done that locally.
> 
> Okay, you're ready to submit the next version?

It's compiling. Then I need to test it. I've also added a couple of
fixes/cleanups I ran into while doing this.