On Fri, Feb 22, 2019 at 05:00:43PM +0800, wenxu@xxxxxxxxx wrote:
> From: wenxu <wenxu@xxxxxxxxx>
>
> The metadata_dst is not init the dst_cache which make the
> ip_md_tunnel_xmit can't use the dst_cache. It will lookup
> route table every packets.
Applied, thanks.
One comment below:
> Signed-off-by: wenxu <wenxu@xxxxxxxxx>
> ---
> net/netfilter/nft_tunnel.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/net/netfilter/nft_tunnel.c b/net/netfilter/nft_tunnel.c
> index ea28588..b113fca 100644
> --- a/net/netfilter/nft_tunnel.c
> +++ b/net/netfilter/nft_tunnel.c
> @@ -406,6 +406,13 @@ static int nft_tunnel_obj_init(const struct nft_ctx *ctx,
> return -ENOMEM;
>
> memcpy(&md->u.tun_info, &info, sizeof(info));
> +#ifdef CONFIG_DST_CACHE
To reduce ifdef pollution, it would be good to turn dst_cache_init()
into noop via static inline trick in case CONFIG_DST_CACHE=n.
BTW, I suspect net/sched/act_tunnel_key.c also needs a similar update.
> + err = dst_cache_init(&md->u.tun_info.dst_cache, GFP_KERNEL);
> + if (err < 0) {
> + metadata_dst_free(md);
> + return err;
> + }
> +#endif
> ip_tunnel_info_opts_set(&md->u.tun_info, &priv->opts.u, priv->opts.len,
> priv->opts.flags);
> priv->md = md;
> --
> 1.8.3.1
>
