On Mon, Apr 06, 2015 at 03:28:37PM -0700, Andy Zhou wrote:
> On Tue, Mar 31, 2015 at 9:52 PM, Ben Pfaff <b...@nicira.com> wrote:
> > To be first used in upcoming commits.
> >
> > Signed-off-by: Ben Pfaff <b...@nicira.com>
>
> > +/* Consider the two value/mask pairs 'a_value/a_mask' and 'b_value/b_mask'
> > as
> > + * restrictions on a field's value. Then, this function initializes
> > + * 'dst_value/dst_mask' such that it combines the restrictions of both
> > pairs.
> > + * This is not always possible, i.e. if one pair insists on a value of 0 in
> > + * some bit and the other pair insists on a value of 1 in that bit. This
> > + * function returns false in a case where the combined restriction is
> > + * impossible (in which case 'dst_value/dst_mask' is not fully
> > initialized),
> > + * true otherwise. */
> > +bool
> > +mf_subvalue_intersect(const union mf_subvalue *a_value,
> > + const union mf_subvalue *a_mask,
> > + const union mf_subvalue *b_value,
> > + const union mf_subvalue *b_mask,
> > + union mf_subvalue *dst_value,
> > + union mf_subvalue *dst_mask)
> > +{
> > + for (int i = 0; i < ARRAY_SIZE(a_value->be64); i++) {
> > + ovs_be64 av = a_value->be64[i];
> > + ovs_be64 am = a_mask->be64[i];
> > + ovs_be64 bv = b_value->be64[i];
> > + ovs_be64 bm = b_mask->be64[i];
> > + ovs_be64 *dv = &dst_value->be64[i];
> > + ovs_be64 *dm = &dst_mask->be64[i];
> > +
> > + if ((av ^ bv) & (am & bm)) {
> > + return false;
> > + }
> > + *dv = av | bv;
> > + *dm = am | bm;
> Should this be *dv = (av & am) | (bv & bm)? I have not read the following
> patch
> to check the actual usage, so my interpretation of 'intersect' may be
> different than
> what's intended here.
I think that in the cases where the expression you suggest would have a
different value, the "if" statement just before would have bailed out.
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
