[dpdk-dev] [PATCH v5 0/3] i40e VXLAN TX checksum rework

[Ananyev, Konstantin]

> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon at 6wind.com]
> Sent: Friday, December 05, 2014 4:07 PM
> To: Liu, Jijiang
> Cc: dev at dpdk.org; Ananyev, Konstantin; olivier.matz at 6wind.com
> Subject: Re: [dpdk-dev] [PATCH v5 0/3] i40e VXLAN TX checksum rework
> 
> > > We have got some feedback about backward compatibility of VXLAN TX checksum offload API with 1G/10G NIC after the i40e
> VXLAN
> > > TX checksum codes were applied, so we have to rework the APIs on i40e, including the changes of mbuf, i40e PMD and csum
> forward
> > > engine.
> > >
> > > The main changes in mbuf are as follows, in place of removing PKT_TX_VXLAN_CKSUM, we introduce 4 new flags:
> > > PKT_TX_OUTER_IP_CKSUM, PKT_TX_OUTER_IPV4, PKT_TX_OUTER_IPV6 and PKT_TX_UDP_TUNNEL_PKT. Replace the
> inner_l2_len
> > > and the inner_l3_len field with the outer_l2_len and outer_l3_len field.
> > >
> > > Let's use a few examples to demonstrate how to use these new flags and existing flags in rte_mbuf.h
> > > Let say we have a tunnel packet: eth_hdr_out/ipv4_hdr_out/udp_hdr_out/vxlan_hdr/ehtr_hdr_in/ipv4_hdr_in/tcp_hdr_in.
> There
> > > could be several scenarios:
> > >
> > > A) User requests HW offload for ipv4_hdr_out checksum.
> > > He doesn't care is it a tunnelled packet or not. So he sets:
> > >
> > > mb->l2_len = eth_hdr_out;
> > > mb->l3_len = ipv4_hdr_out;
> > > mb->ol_flags |= PKT_TX_IPV4_CSUM;
> > >
> > > B) User is aware that it is a tunnelled packet and requests HW offload for ipv4_hdr_in and tcp_hdr_in *only*.
> > > He doesn't care about outer IP checksum offload. In that case, for FVL  he has 2 choices:
> > >    1. Treat that packet as a 'proper' tunnelled packet, and fill all the fields:
> > >      mb->l2_len = udp_hdr_out + vxlan_hdr +eth_hdr_in;
> > >      mb->l3_len = ipv4_hdr_in;
> > >      mb->outer_l2_len = eth_hdr_out;
> > >      mb->outer_l3_len = ipv4_hdr_out;
> > >      mb->ol_flags |= PKT_TX_UDP_TUNNEL_PKT | PKT_TX_IP_CKSUM |  PKT_TX_TCP_CKSUM;
> > >
> > >    2. As user doesn't care about outer IP hdr checksum, he can treat everything before ipv4_hdr_in as L2 header.
> > >    So he knows, that it is a tunnelled packet, but makes HW to treat it as ordinary (non-tunnelled) packet:
> > >      mb->l2_len = eth_hdr_out + ipv4_hdr_out + udp_hdr_out + vxlan_hdr + ehtr_hdr_in;
> > >      mb->l3_len = ipv4_hdr_in;
> > >      mb->ol_flags |= PKT_TX_IP_CKSUM |  PKT_TX_TCP_CKSUM;
> > >
> > > i40e PMD will support both B.1 and B.2, but ixgbe/igb/em PMD supports only B.2.
> > > if HW supports both - it will be up to user app which method to choose.
> > > tespmd will support both methods, and it should be configurable by user which approach to use (cmdline parameter).
> > > So the user can try/test both methods and select an appropriate for him.
> > >
> > > C) User knows that is a tunnelled packet, and wants HW offload for all 3 checksums:  outer IP hdr checksum, inner IP checksum,
> inner
> > > TCP checksum.
> > > Then he has to setup all TX checksum fields:
> > >      mb->l2_len =  udp_hdr_out + vxlan_hdr +eth_hdr_in;;
> > >      mb->l3_len = ipv4_hdr_in;
> > >      mb->outer_l2_len = eth_hdr_out;
> > >      mb->outer_l3_len = ipv4_hdr_out;
> > >      mb->ol_flags |= PKT_TX_OUT_IP_CKSUM  | PKT_TX_UDP_TUNNEL_PKT | PKT_TX_IP_CKSUM |  PKT_TX_TCP_CKSUM;
> > >
> > > Change notes:
> > > v2 changes:
> > >      remove PKT_TX_IP_CKSUM alias.
> > >      add PKT_TX_OUT_IP_CKSUM and PKT_TX_OUTER_IPV6 in rte_get_tx_ol_flag_name.
> > >      spliting mbuf changes into two patches.
> > >      fix MACLEN caculation issue in i40e driver
> > >      fix some issues in csumonly.c
> > >      change cover letter.
> > > v3 changes:
> > >      fix MACLEN caculation issue in i40e driver when non-tunneling packet
> > > v4 changes:
> > >      reorganize patches to avoid compilation to be broken between patches.
> > >      remove l4_tun_len from mbuf structure.
> > >      add PKT_TX_OUTER_IPV4 to indicate no IP checksum offload requirement for tunneling packet.
> > >      change i40e PMD and csum engine due to above changes.
> > >
> > > v5 changes:
> > >      according to Konstantin's comments, optimize process_outer_cksums() in order to avoid setting PKT_TX_OUTER_IPV4 flags for
> the
> > > case when user didn't enable TESTPMD_TX_OFFLOAD_VXLAN_CKSUM
> > >
> > > Jijiang Liu (3):
> > >   Redefine PKT_TX_IPV4, PKT_TX_IPV6 and PKT_TX_VLAN_PKT;
> > >   Replace PKT_TX_VXLAN_CKSUM with PKT_TX_UDP_TUNNEL_PKT, and add 3 TX flags, which are PKT_TX_OUTER_IP_CKSUM,
> > > PKT_TX_OUTER_IPV4 and PKT_TX_OUTER_IPV6,and rework csum forward engine and i40e pmd due to these changes;
> > >   Replace the inner_l2_len and the inner_l3_len field with the outer_l2_len and outer_l3_len field, and rework csum forward
> engine
> > > and i40e pmd due to  these changes;
> >
> > Acked-by: Konstantin Ananyev <konstantin.ananyev at intel.com>
> 
> Applied with small comments changes suggested by Olivier.
> 
> Thanks everyone for finding a consensus.
> It's not easy to accept different of views but we finally did it!
> 
> Lesson learned: every details of an API must be explicited with clear and
> short sentences.

>From my side I would like to thank Olivier for all his tremendous help with that feature.
Konstantin

> 
> Thanks
> --
> Thomas