Re: [PATCH net v2] skbuff: Release nfct refcount on napi stolen or re-used skbs

[Paolo Abeni <pabeni@redhat.com>]

On Mon, 2021-07-05 at 10:49 +0300, Paul Blakey wrote:
> When multiple SKBs are merged to a new skb under napi GRO,
> or SKB is re-used by napi, if nfct was set for them in the
> driver, it will not be released while freeing their stolen
> head state or on re-use.
> 
> Release nfct on napi's stolen or re-used SKBs, and
> in gro_list_prepare, check conntrack metadata diff.
> 
> Fixes: 5c6b94604744 ("net/mlx5e: CT: Handle misses after executing CT action")
> Reviewed-by: Roi Dayan <roid@nvidia.com>
> Signed-off-by: Paul Blakey <paulb@nvidia.com>
> ---
> Changelog:
> 	v1->v2:
> 	 Check for different flows based on CT and chain metadata in gro_list_prepare
> 
>  net/core/dev.c    | 13 +++++++++++++
>  net/core/skbuff.c |  1 +
>  2 files changed, 14 insertions(+)
> 
> diff --git a/net/core/dev.c b/net/core/dev.c
> index 439faadab0c2..bf62cb2ec6da 100644
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -5981,6 +5981,18 @@ static void gro_list_prepare(const struct list_head *head,
>  			diffs = memcmp(skb_mac_header(p),
>  				       skb_mac_header(skb),
>  				       maclen);
> +
> +		diffs |= skb_get_nfct(p) ^ skb_get_nfct(skb);
> +
> +		if (!diffs) {
> +			struct tc_skb_ext *skb_ext = skb_ext_find(skb, TC_SKB_EXT);
> +			struct tc_skb_ext *p_ext = skb_ext_find(p, TC_SKB_EXT);
> +
> +			diffs |= (!!p_ext) ^ (!!skb_ext);
> +			if (!diffs && unlikely(skb_ext))
> +				diffs |= p_ext->chain ^ skb_ext->chain;
> +		}

I'm wondering... if 2 skbs are merged, and have the same L2/L3/L4
headers - except len and csum - can they have different dst/TC_EXT?

@Eric: I'm sorry for the very dumb and late question. You reported v1
of this patch would make "GRO slow as hell", could you please elaborate
a bit more? I thought most skbs (with no ct attached) would see little
difference???

Cheers,

Paolo