On Thu, 6 Jun 2024 07:58:50 -0700 Vadim Fedorenko wrote:
> @@ -1060,9 +1062,19 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
> __skb_push(skb, hh_len);
> if (is_direct_pkt_access)
> bpf_compute_data_pointers(skb);
> +
> ret = convert___skb_to_skb(skb, ctx);
> if (ret)
> goto out;
> +
> + if (kattr->test.flags & BPF_F_TEST_SKB_CHECKSUM_COMPLETE) {
> + const int off = skb_network_offset(skb);
> + int len = skb->len - off;
> +
> + skb->csum = skb_checksum(skb, off, len, 0);
> + skb->ip_summed = CHECKSUM_COMPLETE;
> + }
Looks good, overall, although I'd be tempted to place this before
the L2 is pushed, a few lines up, so that we don't need to worry
about network offset. Then again, with you approach there is a nice
symmetry between the pre- and post- if blocks so either way is fine:
Reviewed-by: Jakub Kicinski <kuba@xxxxxxxxxx>
