Re: [PATCH RFC bpf-next v2 10/17] bnxt: Propagate trait presence to skb

[Stanislav Fomichev <stfomichev@gmail.com>]

On 04/22, Arthur Fabre wrote:
> Call the common xdp_buff_update_skb() helper.
> 
> Signed-off-by: Arthur Fabre <arthur@arthurfabre.com>
> ---
>  drivers/net/ethernet/broadcom/bnxt/bnxt.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
> index c8e3468eee612ad622bfbecfd7cc1ae3396061fd..0eba3e307a3edbc5fe1abf2fa45e6256d98574c2 100644
> --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
> +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
> @@ -2297,6 +2297,10 @@ static int bnxt_rx_pkt(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
>  			}
>  		}
>  	}
> +
> +	if (xdp_active)
> +		xdp_buff_update_skb(&xdp, skb);

For me, the preference for reusing existing metadata area was
because of the patches 10-16: we now need to care about two types of
metadata explicitly.

If you insist on placing it into the headroom, can we at least have some
common helper to finish xdp->skb conversion? It can call skb_ext_from_headroom
and/or skb_metadata_set:

xdp_buff_done(*xdp, *skb) {
	if (have traits) {
		skb_ext_from_headroom
		return
	}

	metasize = xdp->data - xdp->data_meta;
	if (metasize)
		skb_metadata_set
}

And then we'll have some common rules for the drivers: call xdp_buff_done
when you're done with the xdp_buff to take care of metadata/traits. And
it might be easier to review: you're gonna (mostly) change existing
calls to skb_metadata_set to your new helper. Maybe we can even
eventually fold all xdp_update_skb_shared_info stuff into that as
well...