[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CANn89i+rN=pcGquBONMruWUS8WG708rqMKD7TjS0OkBq+ev+Ew@mail.gmail.com>
Date: Wed, 1 Feb 2023 09:53:45 +0100
From: Eric Dumazet <edumazet@...gle.com>
To: Xin Long <lucien.xin@...il.com>
Cc: network dev <netdev@...r.kernel.org>, davem@...emloft.net,
kuba@...nel.org, Paolo Abeni <pabeni@...hat.com>,
David Ahern <dsahern@...il.com>,
Hideaki YOSHIFUJI <yoshfuji@...ux-ipv6.org>,
Pravin B Shelar <pshelar@....org>,
Jamal Hadi Salim <jhs@...atatu.com>,
Cong Wang <xiyou.wangcong@...il.com>,
Jiri Pirko <jiri@...nulli.us>,
Pablo Neira Ayuso <pablo@...filter.org>,
Florian Westphal <fw@...len.de>,
Marcelo Ricardo Leitner <marcelo.leitner@...il.com>,
Ilya Maximets <i.maximets@....org>,
Aaron Conole <aconole@...hat.com>,
Roopa Prabhu <roopa@...dia.com>,
Nikolay Aleksandrov <razor@...ckwall.org>,
Mahesh Bandewar <maheshb@...gle.com>,
Paul Moore <paul@...l-moore.com>,
Guillaume Nault <gnault@...hat.com>
Subject: Re: [PATCHv4 net-next 00/10] net: support ipv4 big tcp
On Sat, Jan 28, 2023 at 4:58 PM Xin Long <lucien.xin@...il.com> wrote:
>
> This is similar to the BIG TCP patchset added by Eric for IPv6:
>
> https://lwn.net/Articles/895398/
>
> Different from IPv6, IPv4 tot_len is 16-bit long only, and IPv4 header
> doesn't have exthdrs(options) for the BIG TCP packets' length. To make
> it simple, as David and Paolo suggested, we set IPv4 tot_len to 0 to
> indicate this might be a BIG TCP packet and use skb->len as the real
> IPv4 total length.
>
> This will work safely, as all BIG TCP packets are GSO/GRO packets and
> processed on the same host as they were created; There is no padding
> in GSO/GRO packets, and skb->len - network_offset is exactly the IPv4
> packet total length; Also, before implementing the feature, all those
> places that may get iph tot_len from BIG TCP packets are taken care
> with some new APIs:
>
> Patch 1 adds some APIs for iph tot_len setting and getting, which are
> used in all these places where IPv4 BIG TCP packets may reach in Patch
> 2-7, Patch 8 adds a GSO_TCP tp_status for af_packet users, and Patch 9
> add new netlink attributes to make IPv4 BIG TCP independent from IPv6
> BIG TCP on configuration, and Patch 10 implements this feature.
>
> Note that the similar change as in Patch 2-6 are also needed for IPv6
> BIG TCP packets, and will be addressed in another patchset.
>
> The similar performance test is done for IPv4 BIG TCP with 25Gbit NIC
> and 1.5K MTU:
>
> No BIG TCP:
> for i in {1..10}; do netperf -t TCP_RR -H 192.168.100.1 -- -r80000,80000 -O MIN_LATENCY,P90_LATENCY,P99_LATENCY,THROUGHPUT|tail -1; done
> 168 322 337 3776.49
> 143 236 277 4654.67
> 128 258 288 4772.83
> 171 229 278 4645.77
> 175 228 243 4678.93
> 149 239 279 4599.86
> 164 234 268 4606.94
> 155 276 289 4235.82
> 180 255 268 4418.95
> 168 241 249 4417.82
>
> Enable BIG TCP:
> ip link set dev ens1f0np0 gro_ipv4_max_size 128000 gso_ipv4_max_size 128000
> for i in {1..10}; do netperf -t TCP_RR -H 192.168.100.1 -- -r80000,80000 -O MIN_LATENCY,P90_LATENCY,P99_LATENCY,THROUGHPUT|tail -1; done
> 161 241 252 4821.73
> 174 205 217 5098.28
> 167 208 220 5001.43
> 164 228 249 4883.98
> 150 233 249 4914.90
> 180 233 244 4819.66
> 154 208 219 5004.92
> 157 209 247 4999.78
> 160 218 246 4842.31
> 174 206 217 5080.99
>
> Thanks for the feedback from Eric and David Ahern.
>
> v1->v2:
> - remove the fixes and the selftest for IPv6 BIG TCP, will do it in
> another patchset.
> - add GSO_TCP for tp_status in packet sockets to tell the af_packet
> users that this is a TCP GSO packet in Patch 8.
> - also check skb_is_gso() when checking if it's a GSO TCP packet in
> Patch 1.
> v2->v3:
> - add gso/gro_ipv4_max_size per device and netlink attributes for them
> in Patch 9, so that we can selectively enable BIG TCP for IPv6, and
> not for IPv4, as Eric required.
> - remove the selftest, as it requires userspace iproute2 change after
> making IPv4 BIG TCP independent from IPv6 BIG TCP on configuration.
> v3->v4:
> - put gso/gro_ipv4_max_size close to other related fields, so that we
> do not need an extra cache line miss, as Eric suggested.
> - also check ipv6_addr_v4mapped() when reading gso(_ipv4)_max_size in
> sk_setup_caps(), as Eric noticed.
For the series:
Reviewed-by: Eric Dumazet <edumazet@...gle.com>
Please make sure to add needed changes to tcpdump/libpcap
Powered by blists - more mailing lists