On Mon, 09 Jun 2025 09:50:30 -0400 Willem de Bruijn wrote: > > > + if (ret) > > > + return ret; > > > + > > > bpf_compute_data_pointers(skb); > > > - return ret; > > I wonder whether that unconditional call to bpf_compute_data_pointers > even if ret was there for a reason. > > From reviewing the bpf_skb_proto_xlat error paths, it does seem safe > to remove it. The cases where an error may be returned after the skb > is modified only modify the skb in terms of headroom, not headlen. I should have mentioned, I looked around and also concluded this unconditional recompute was purely aesthetic.
