On Wed, Jun 18, 2014 at 12:17:55AM -0700, Joe Perches wrote:
> On Wed, 2014-06-18 at 00:06 -0700, Greg Kroah-Hartman wrote:
> > On Wed, Jun 18, 2014 at 09:49:39AM +0300, Alexey Tulia wrote:
> > > On Tue, Jun 17, 2014 at 03:44:58PM -0700, Greg Kroah-Hartman wrote:
> > > > On Fri, Jun 13, 2014 at 11:35:13AM +0300, Alexey Tulia wrote:
> > > > > This fixes the following warning:
> > > > > - WARNING: __constant_cpu_to_le32 should be cpu_to_le32
> > > >
> > > > What produces this warning?
> > > >
> > >
> > > This warning was found by checkpatch.pl -f
> > >
> > > > >
> > > > > Signed-off-by: Alexey Tulia <alexey.tulia@xxxxxxxxx>
> > > > > ---
> > > > > drivers/staging/usbip/vhci_hcd.c | 2 +-
> > > > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > > > >
> > > > > diff --git a/drivers/staging/usbip/vhci_hcd.c b/drivers/staging/usbip/vhci_hcd.c
> > > > > index 0007d30..e21c1b4 100644
> > > > > --- a/drivers/staging/usbip/vhci_hcd.c
> > > > > +++ b/drivers/staging/usbip/vhci_hcd.c
> > > > > @@ -304,7 +304,7 @@ static int vhci_hub_control(struct usb_hcd *hcd, u16 typeReq, u16 wValue,
> > > > > break;
> > > > > case GetHubStatus:
> > > > > usbip_dbg_vhci_rh(" GetHubStatus\n");
> > > > > - *(__le32 *) buf = __constant_cpu_to_le32(0);
> > > > > + *(__le32 *) buf = cpu_to_le32(0);
> > > >
> > > > How is this correct? Why shouldn't __constant_cpu_to_le32() be used
> > > > here? Heck, why can't we just use 0 given that it doesn't matter the
> > > > endianness of that specific value :)
> > >
> > > It may be so, but anyway the __constant_cpu_to_le32 produced a warning
> > > and it was obvious to clean up this part of code a bit.
> > > However, the 0 value possibly can change to other value in future and
> > > this macro acts as a safety net here.
> >
> > Dragging Joe in here as he wrote that checkpatch feature.
> >
> > Joe, how is not using __constant_cpu_to_* a good thing? If I have a
> > constant value, cpu_to_* will be a function call if the bits have to be
> > switched around (no-op if not). I don't see the code path that detects
> > a constant value and calls the swap-bits-as-a-constant macro instead,
> > unless the __constant_cpu_to* function is called.
> >
> > Am I just missing it somewhere in the .h include chain?
>
> Probably.
>
> There a __builtin_constant_p check in there via the
> include/uapi/linux/swab.h chain.
>
> for instance:
>
> htonl ->
> ___htonl ->
> __cpu_to_be32
> __constant_htonl ->
> __swab32 ->
>
> include/uapi/linux/swab.h:#define __swab32(x) \
> include/uapi/linux/swab.h- (__builtin_constant_p((__u32)(x)) ? \
> include/uapi/linux/swab.h- ___constant_swab32(x) : \
> include/uapi/linux/swab.h- __fswab32(x))
Ah, ok, my fault, sorry for the noise. I'll go queue this patch up in a
bit...
greg k-h
--
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
