Re: [RFT PATCH 7/9] ethtool: prepare for larger netdev_features_t type

[Ben Hutchings <bhutchings@solarflare.com>]

On Thu, 2011-06-23 at 10:50 -0700, Mahesh Bandewar wrote:
> On Mon, Jun 20, 2011 at 2:16 PM, Ben Hutchings
> <bhutchings@solarflare.com> wrote:
> >
> > On Mon, 2011-06-20 at 21:14 +0200, Michał Mirosław wrote:
> > [...]
> > > @@ -125,19 +131,26 @@ static int ethtool_set_features(struct net_device *dev, void __user *useraddr)
> > >       if (copy_from_user(features, useraddr, sizeof(features)))
> > >               return -EFAULT;
> > >
> > > -     if (features[0].valid & ~NETIF_F_ETHTOOL_BITS)
> > > +     /* I wonder if the compiler will be smart enough to loop-unroll
> > > +      * and optimize this... (no worries if not) --mq */
> > > +     for (i = ETHTOOL_DEV_FEATURE_WORDS; i-- > 0; ) {
> > > +             valid = (valid << 32)|features[i].valid;
> > > +             wanted = (wanted << 32)|features[i].requested;
> > > +     }
> > [...]
> >
> > I don't know (or care) about optimisation of this, but I would expect
> > gcc to complain about shifting a 32-bit value by 32 bits.  I suggest you
> > write this as:
> >
> >        for (i = 0; i < ETHTOOL_DEV_FEATURE_WORDS; ++i) {
> >                valid |= (netdev_features_t)features[i].valid << 32 *i;
> >                wanted |= (netdev_features_t)features[i].requested << 32 *i;
> 
> It's a valid point but this type of typecast or similar usage would
> imply that netdev_feature_t is an int of XXX bits. That's not opaque
> and would hinder the way you can abstract the feature type.

Yes, ethtool_{get,set}_features() will have to be changed if and when
the representation of netdev_features_t is changed significantly.  I
don't think there's any way of avoiding that and I don't think it really
matters.

Ben.

-- 
Ben Hutchings, Senior Software Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.