Re: [PATCH v3] net: fix wrong network header length

From: Paolo Abeni <pabeni@redhat.com>
To: Lina Wang <lina.wang@mediatek.com>,
	"David S . Miller" <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>,
	Matthias Brugger <matthias.bgg@gmail.com>,
	Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Andrii Nakryiko <andrii@kernel.org>
Cc: "Network Development" <netdev@vger.kernel.org>,
	LKML <linux-kernel@vger.kernel.org>,
	linux-arm-kernel <linux-arm-kernel@lists.infradead.org>,
	bpf <bpf@vger.kernel.org>,
	"Maciej Żenczykowski" <maze@google.com>,
	"Willem de Bruijn" <willemb@google.com>,
	"Eric Dumazet" <edumazet@google.com>
Subject: Re: [PATCH v3] net: fix wrong network header length
Date: Thu, 17 Feb 2022 09:45:27 +0100	[thread overview]
Message-ID: <41a3e6e4f9331c6a0a62fe838fc6f9084a5c89bc.camel@redhat.com> (raw)
In-Reply-To: <20220217070139.30028-1-lina.wang@mediatek.com>

Hello,

On Thu, 2022-02-17 at 15:01 +0800, Lina Wang wrote:
> On Wed, 2022-02-16 at 19:05 -0800, Alexei Starovoitov wrote:
> > On Tue, Feb 15, 2022 at 11:37 PM Lina Wang <lina.wang@mediatek.com>
> > wrote:
> > > 
> > > When clatd starts with ebpf offloaing, and NETIF_F_GRO_FRAGLIST is
> > > enable,
> > > several skbs are gathered in skb_shinfo(skb)->frag_list. The first
> > > skb's
> > > ipv6 header will be changed to ipv4 after bpf_skb_proto_6_to_4,
> > > network_header\transport_header\mac_header have been updated as
> > > ipv4 acts,
> > > but other skbs in frag_list didnot update anything, just ipv6
> > > packets.
> > 
> > Please add a test that demonstrates the issue and verifies the fix.
> 
> I used iperf udp test to verify the patch, server peer enabled -d to debug
> received packets.
> 
> 192.0.0.4 is clatd interface ip, corresponding ipv6 addr is 
> 2000:1:1:1:afca:1b1f:1a9:b367, server peer ip is 1.1.1.1,
> whose ipv6 is 2004:1:1:1::101:101.
> 
> Without the patch, when udp length 2840 packets received, iperf shows:
> pcount 1 packet_count 0
> pcount 27898727 packet_count 1
> pcount 3 packet_count 27898727
> 
> pcount should be 2, but is 27898727(0x1a9b367) , which is 20 bytes put 
> forward. 
> 
> 12:08:02.680299	Unicast to us 2004:1:1:1::101:101   2000:1:1:1:afca:1b1f:1a9:b367 UDP 51196 → 5201 Len=2840
> 0000   20 00 00 01 00 01 00 01 af ca 1b 1f 01 a9 b3 67   ipv6 dst address
> 0000   c7 fc 14 51 0b 20 c7 ab                           udp header
> 0000   00 00 00 ab 00 0e f3 49 00 00 00 01 08 06 69 d2   00000001 is pcount
> 12:08:02.682084	Unicast to us	1.1.1.1	                 192.0.0.4 	 	  UDP 51196 → 5201 Len=2840
> 
> After applied the patch, there is no OOO, pcount acted in order.

To clarify: Alexei is asking to add a test under:

tools/testing/selftests/net/

to cover this specific case. You can propbably extend the existing
udpgro_fwd.sh.

Please explicitly CC people who gave feedback to previous iterations,
it makes easier to track the discussion.

/P