Received-SPF: pass (google.com: domain of linux-kernel+bounces-159657-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.48.161 as permitted sender) client-ip=147.75.48.161;
Precedence: bulk
MIME-Version: 1.0
References: <20240426065143.4667-1-nbd@nbd.name> <20240426065143.4667-4-nbd@nbd.name>
In-Reply-To: <20240426065143.4667-4-nbd@nbd.name>
From: Eric Dumazet <edumazet@google.com>
Date: Fri, 26 Apr 2024 09:47:30 +0200
Message-ID: <CANn89iL-4CB3-+mMtBQapXV7EaCzOWOU1MRp4X1tUtjaXfbk=Q@mail.gmail.com>
Subject: Re: [PATCH v3 net-next v3 3/6] net: add code for TCP fraglist GRO
To: Felix Fietkau <nbd@nbd.name>
Cc: netdev@vger.kernel.org, "David S. Miller" <davem@davemloft.net>, 
	David Ahern <dsahern@kernel.org>, Jakub Kicinski <kuba@kernel.org>, Paolo Abeni <pabeni@redhat.com>, 
	willemdebruijn.kernel@gmail.com, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Fri, Apr 26, 2024 at 8:51=E2=80=AFAM Felix Fietkau <nbd@nbd.name> wrote:
>
> This implements fraglist GRO similar to how it's handled in UDP, however
> no functional changes are added yet. The next change adds a heuristic for
> using fraglist GRO instead of regular GRO.
>
> Signed-off-by: Felix Fietkau <nbd@nbd.name>
> ---
>  net/ipv4/tcp_offload.c   | 22 ++++++++++++++++++++++
>  net/ipv6/tcpv6_offload.c |  9 +++++++++
>  2 files changed, 31 insertions(+)
>
> diff --git a/net/ipv4/tcp_offload.c b/net/ipv4/tcp_offload.c
> index c493e95e09a5..ffd6b7a4163a 100644
> --- a/net/ipv4/tcp_offload.c
> +++ b/net/ipv4/tcp_offload.c
> @@ -332,6 +332,19 @@ struct sk_buff *tcp_gro_receive(struct list_head *he=
ad, struct sk_buff *skb)
>         flush |=3D (ntohl(th2->seq) + skb_gro_len(p)) ^ ntohl(th->seq);
>         flush |=3D skb_cmp_decrypted(p, skb);
>
> +       if (NAPI_GRO_CB(p)->is_flist) {


Please add unlikely() for all NAPI_GRO_CB(p)->is_flist checks added in
this patch.

> +               flush |=3D (__force int)(flags ^ tcp_flag_word(th2));
> +               flush |=3D skb->ip_summed !=3D p->ip_summed;
> +               flush |=3D skb->csum_level !=3D p->csum_level;
> +               flush |=3D !pskb_may_pull(skb, skb_gro_offset(skb));
> +               flush |=3D NAPI_GRO_CB(p)->count >=3D 64;
> +
> +               if (flush || skb_gro_receive_list(p, skb))
> +                       mss =3D 1;
> +
> +               goto out_check_final;
> +       }
> +
>         if (flush || skb_gro_receive(p, skb)) {
>                 mss =3D 1;
>                 goto out_check_final;
> @@ -398,6 +411,15 @@ INDIRECT_CALLABLE_SCOPE int tcp4_gro_complete(struct=
 sk_buff *skb, int thoff)
>         const struct iphdr *iph =3D ip_hdr(skb);
>         struct tcphdr *th =3D tcp_hdr(skb);
>
> +       if (NAPI_GRO_CB(skb)->is_flist) {
> +               skb_shinfo(skb)->gso_type |=3D SKB_GSO_FRAGLIST | SKB_GSO=
_TCPV4;
> +               skb_shinfo(skb)->gso_segs =3D NAPI_GRO_CB(skb)->count;
> +
> +               __skb_incr_checksum_unnecessary(skb);
> +
> +               return 0;
> +       }
> +
>         th->check =3D ~tcp_v4_check(skb->len - thoff, iph->saddr,
>                                   iph->daddr, 0);
>
> diff --git a/net/ipv6/tcpv6_offload.c b/net/ipv6/tcpv6_offload.c
> index b3b8e1f6b92a..c97d55cf036f 100644
> --- a/net/ipv6/tcpv6_offload.c
> +++ b/net/ipv6/tcpv6_offload.c
> @@ -32,6 +32,15 @@ INDIRECT_CALLABLE_SCOPE int tcp6_gro_complete(struct s=
k_buff *skb, int thoff)
>         const struct ipv6hdr *iph =3D ipv6_hdr(skb);
>         struct tcphdr *th =3D tcp_hdr(skb);
>
> +       if (NAPI_GRO_CB(skb)->is_flist) {
> +               skb_shinfo(skb)->gso_type |=3D SKB_GSO_FRAGLIST | SKB_GSO=
_TCPV6;
> +               skb_shinfo(skb)->gso_segs =3D NAPI_GRO_CB(skb)->count;
> +
> +               __skb_incr_checksum_unnecessary(skb);
> +
> +               return 0;
> +       }
> +
>         th->check =3D ~tcp_v6_check(skb->len - thoff, &iph->saddr,
>                                   &iph->daddr, 0);
>         skb_shinfo(skb)->gso_type |=3D SKB_GSO_TCPV6;
> --
> 2.44.0
>