Re: [PATCH] usb: host: fhci-hcd: annotate PIPE_CONTROL switch case with fallthrough

[Greg Kroah-Hartman]

On Thu, Feb 13, 2020 at 02:35:18PM +0100, Rasmus Villemoes wrote:
> On 13/02/2020 13.56, Greg Kroah-Hartman wrote:
> 
> >> diff --git a/drivers/usb/host/fhci-hcd.c b/drivers/usb/host/fhci-hcd.c
> >> index 04733876c9c6..a8e1048278d0 100644
> >> --- a/drivers/usb/host/fhci-hcd.c
> >> +++ b/drivers/usb/host/fhci-hcd.c
> >> @@ -396,6 +396,7 @@ static int fhci_urb_enqueue(struct usb_hcd *hcd, struct urb *urb,
> >>  	case PIPE_CONTROL:
> >>  		/* 1 td fro setup,1 for ack */
> >>  		size = 2;
> >> +		fallthrough;
> > 
> > We have an attribute for that?
> > 
> > Shouldn't this be /* fall through */ instead?
> > 
> > Gustavo, what's the best practice here, I count only a few
> > "fallthrough;" instances in the kernel, although one is in our coding
> > style document, and thousands of the /* */ version.
> 
> Yes, I went with the attribute/macro due to that, and the history is
> that Linus applied Joe's patches directly
> (https://lore.kernel.org/lkml/CAHk-=whOF8heTGz5tfzYUBp_UQQzSWNJ_50M7-ECXkfFRDQWFA@xxxxxxxxxxxxxx/),
> so I assumed that meant the Penguin decided that the attribute/macro is
> the right thing to do for new code, while existing comment annotations
> can be left alone or changed piecemeal as code gets refactored anyway.

But, to be fair, Gustavo went and fixed up thousands of these, with the
/* */ version, not the attribute.

Gustavo, can coverity notice the "fallthrough;" attribute properly?  I
don't want to start adding things that end up triggering
false-positives.

thanks,

greg k-h