On Tue, Jun 18, 2019 at 10:09:51AM -0300, Jason Gunthorpe wrote:
> On Tue, Jun 18, 2019 at 01:04:14PM +0000, Leon Romanovsky wrote:
> > On Tue, Jun 18, 2019 at 10:01:50AM -0300, Jason Gunthorpe wrote:
> > > On Tue, Jun 18, 2019 at 12:19:04PM +0000, Leon Romanovsky wrote:
> > > > > diff --git a/include/uapi/rdma/rdma_netlink.h b/include/uapi/rdma/rdma_netlink.h
> > > > > index 9903db21a42c58..b27c02185dcc19 100644
> > > > > +++ b/include/uapi/rdma/rdma_netlink.h
> > > > > @@ -504,6 +504,7 @@ enum rdma_nldev_attr {
> > > > > RDMA_NLDEV_ATTR_CHARDEV_NAME, /* string */
> > > > > RDMA_NLDEV_ATTR_CHARDEV_ABI, /* u64 */
> > > > > RDMA_NLDEV_ATTR_CHARDEV, /* u64 */
> > > > > + RDMA_NLDEV_ATTR_UVERBS_DRIVER_ID, /* u64 */
> > > >
> > > > This should be inside nla_policy too.
> > >
> > > It is an output, not an input. policy only checks inputs.
> >
> > We are putting in policy everything to ensure that it won't be forgotten
> > once output field will be used as an input.
>
> Adding dead never tested code is more likely to just get it wrong..
What are you talking? It is addition of extra line in the table, which
will be enforced during first attempt to use it as an input.
Less chances to be broken after.
Thanks
>
> Jason
