Re: [Xen-devel] [PATCH] grant-table, xen-netback: Introduce helper functions for grant copy operations

[Ian Campbell <Ian.Campbell@citrix.com>]

On Thu, 2014-04-03 at 10:48 +0100, David Vrabel wrote:
> On 03/04/14 09:12, Paul Durrant wrote:
> > Zoltan Kiss wrote:
> >>
> >> Create helper functions for grant copy operations and use them in netback.
> >>
> [...]
> >> --- a/drivers/net/xen-netback/netback.c
> >> +++ b/drivers/net/xen-netback/netback.c
> >> @@ -275,23 +275,29 @@ static void xenvif_gop_frag_copy(struct xenvif *vif,
> >> struct sk_buff *skb,
> >>  			bytes = MAX_BUFFER_OFFSET - npo->copy_off;
> >>
> >>  		copy_gop = npo->copy + npo->copy_prod++;
> >> -		copy_gop->flags = GNTCOPY_dest_gref;
> >> -		copy_gop->len = bytes;
> >>
> >>  		if (foreign_vif) {
> >> -			copy_gop->source.domid = foreign_vif->domid;
> >> -			copy_gop->source.u.ref = foreign_gref;
> >> -			copy_gop->flags |= GNTCOPY_source_gref;
> >> +			gnttab_set_copy_op_ref_to_ref(copy_gop,
> >> +						      foreign_gref,
> >> +						      npo->copy_gref,
> >> +						      foreign_vif->domid,
> >> +						      offset,
> >> +						      vif->domid,
> >> +						      npo->copy_off,
> >> +						      bytes,
> >> +						      GNTCOPY_dest_gref |
> >> +						      GNTCOPY_source_gref);
> > 
> > Can't you lose the flags here (and in the other variants)? Since
> > they are implied by the variant of the function they could just be hardcoded
> > there, couldn't they?
> 
> Even with that change these are still functions with 7 or 8 parameters
> with no obvious order.  That's just awful.  The open-coded struct
> initialization is better.

Yes, I think this was a failed experiment.

The original code would likely still be improvable by using a temporary
variable to hold "vif->tx_copy_ops[*copy_ops]", basically the way the
mapping ops are via mop->.

Ian.