Hi, Brian, Thanks indeed for the quick response. Network adaptor is X552 which integrated in XEON, https://www.intel.com/content/dam/www/public/us/en/documents/datasheets/ethernet-x550-datasheet.pdf
The MTU in PF is “mtu 9194” 3: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9194 qdisc mq state UP mode DEFAULT group default qlen 1000 In current test, I sent the packet size is 1kB, maybe small packets. OK, thanks, I’ll also try set dpdk mtu, to avoid big packets drop in HW filtering. Is it like this? rte_eth_dev_set_mtu(port_id, 9194 - 8). Best Regards, Dave(Jianyue) From: Johnson, Brian <[email protected]> Sent: 2020年7月14日 10:53 To: Wu, Jianyue (NSB - CN/Hangzhou) <[email protected]>; Li, Xiaoyun <[email protected]>; [email protected] Cc: Harouat, Karim (Nokia - FR/Lannion) <[email protected]>; Meunier, Julien (Nokia - FR/Paris-Saclay) <[email protected]>; Cheng, Fei 2. (NSB - CN/Hangzhou) <[email protected]>; Lu, Malak (NSB - CN/Hangzhou) <[email protected]>; Li, Jiu (NSB - CN/Hangzhou) <[email protected]>; Pan, Jianyong (NSB - CN/Hangzhou) <[email protected]>; Hong, Christian (NSB - CN/Hangzhou) <[email protected]>; Fei, Xiang (NSB - CN/Hangzhou) <[email protected]>; Ning, Gang (NSB - CN/Hangzhou) <[email protected]>; Ren, Longlong (NSB - CN/Hangzhou) <[email protected]>; Ye, Hua (NSB - CN/Hangzhou) <[email protected]> Subject: Re: [dpdk-users] [dpdk-18.05] Issue while configuring VLAN ID and priority in VF What network adapter are you using? In some cases if there is a port vlan set, since the VF driver does not know to add an additional 4B to the max Rx packet length, the packets get dropped by the hardware filters if the packet is at full MTU size. Try setting the MTU in DPDK and this should add the 4B or 8B needed. ________________________________ From: users <[email protected]<mailto:[email protected]>> on behalf of Wu, Jianyue (NSB - CN/Hangzhou) <[email protected]<mailto:[email protected]>> Sent: Monday, July 13, 2020 7:43:38 PM To: Li, Xiaoyun <[email protected]<mailto:[email protected]>>; [email protected]<mailto:[email protected]> <[email protected]<mailto:[email protected]>> Cc: Harouat, Karim (Nokia - FR/Lannion) <[email protected]<mailto:[email protected]>>; Meunier, Julien (Nokia - FR/Paris-Saclay) <[email protected]<mailto:[email protected]>>; Cheng, Fei 2. (NSB - CN/Hangzhou) <[email protected]<mailto:[email protected]>>; Lu, Malak (NSB - CN/Hangzhou) <[email protected]<mailto:[email protected]>>; Li, Jiu (NSB - CN/Hangzhou) <[email protected]<mailto:[email protected]>>; Pan, Jianyong (NSB - CN/Hangzhou) <[email protected]<mailto:[email protected]>>; Hong, Christian (NSB - CN/Hangzhou) <[email protected]<mailto:[email protected]>>; Fei, Xiang (NSB - CN/Hangzhou) <[email protected]<mailto:[email protected]>>; Ning, Gang (NSB - CN/Hangzhou) <[email protected]<mailto:[email protected]>>; Ren, Longlong (NSB - CN/Hangzhou) <[email protected]<mailto:[email protected]>>; Ye, Hua (NSB - CN/Hangzhou) <[email protected]<mailto:[email protected]>> Subject: [dpdk-users] [dpdk-18.05] Issue while configuring VLAN ID and priority in VF Hi, xiaoyun, Could you help give a hand about this issue? Thanks indeed:) Background: Simplified traffic flow is like this: Sender XEON X552 eth1 vf 0 -> Receiver XEON X552 eth1 vf 0. Our requirement is to configure DPDK to support one VLAN ID, and with multiple priorities 0~7. Current configuration is: PVID: rte_eth_dev_set_vlan_pvid(port_id, 201, 1); VLAN offload: vlan_offload = rte_eth_dev_get_vlan_offload(port_id); vlan_offload &= ~ETH_VLAN_STRIP_OFFLOAD; vlan_offload &= ~ETH_VLAN_FILTER_OFFLOAD; vlan_offload &= ~ETH_VLAN_EXTEND_OFFLOAD; if (0 != rte_eth_dev_set_vlan_offload(port_id, vlan_offload)) Spoof off, trust on: vf 0 MAC 02:40:43:80:01:39, spoof checking off, link-state auto, trust on, query_rss off Test scenarios: ==================================================== First scenario: set vlan in vf 0 -- priority always 0 ==================================================== ip link show eth1 3: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9194 qdisc mq state UP mode DEFAULT group default qlen 1000 link/ether 02:40:43:80:01:38 brd ff:ff:ff:ff:ff:ff vf 0 MAC 02:40:43:80:01:39, vlan 201, spoof checking off, link-state auto, trust on, query_rss off If we configure vlan in vf 0, then we can only send out packets with priority 0. Even we already enabled mbuf->ol_flags |= PKT_TX_VLAN_PKT; and set correct mbuf->vlan_tci. ====================================================== Second scenario: not set vlan in vf 0 -- can't recv ====================================================== 3: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9194 qdisc mq state UP mode DEFAULT group default qlen 1000 link/ether 02:40:43:80:01:38 brd ff:ff:ff:ff:ff:ff vf 0 MAC 02:40:43:80:01:39, spoof checking off, link-state auto, trust on, query_rss off If we don't set vlan in vf 0, then we can send out in sender XEON X552 eth1 vf 0 with correct VLAN ID and priority, From tcpdump in mirrored traffic of receiver XEON, we can see the packet has correct VLAN ID and priority, but the problem is packet silently dropped. From DPDK counters, we can't see the packet received in receiver XEON X552 eth1 vf0. When we enable promiscuous mode, we can see the packets arrive in receiver XEON, and VLAN ID, priority is correct. but can't arrive at DPDK. From my simple assumption, maybe we need second scenario, to support multiple priorities. But how to configure to let the packet pass vf 0? Did I miss something in my configuration? Thanks indeed in advance! Best Regards, Dave(Jianyue)
