Hello, one of our servers crashes regularly, apparently during heavy network load. The log files are then full of this message:
kernel: [514257.305733] i40e 0000:02:00.0: AMD-Vi: Event logged [IO_PAGE_FAULT domain=0x0020 address=0x79ea8113f60 flags=0x0000] This is the driver: i40e: Intel(R) Ethernet Connection XL710 Network Driver i40e: Copyright (c) 2013 - 2019 Intel Corporation. i40e 0000:02:00.0: fw 8.5.67516 api 1.15 nvm 8.50 0x8000be1e 1.3295.0 [8086:15ff] [15d9:1c76] i40e 0000:02:00.0: MAC address: 7c:c2:55:9d:d2:78 i40e 0000:02:00.0: FW LLDP is enabled i40e 0000:02:00.0 eth0: NIC Link is Up, 10 Gbps Full Duplex, Flow Control: None i40e 0000:02:00.0: PCI-Express: Speed 8.0GT/s Width x4 i40e 0000:02:00.0: PCI-Express bandwidth available for this device may be insufficient for optimal performance. i40e 0000:02:00.0: Please move the device to a different PCI-e link with more lanes and/or higher transfer rate. i40e 0000:02:00.0: Features: PF-id[0] VFs: 64 VSIs: 66 QP: 119 RSS FD_ATR FD_SB NTUPLE DCB VxLAN Geneve PTP VEPA i40e 0000:02:00.1: fw 8.5.67516 api 1.15 nvm 8.50 0x8000be1e 1.3295.0 [8086:15ff] [15d9:1c76] i40e 0000:02:00.1: MAC address: 7c:c2:55:9d:d2:79 i40e 0000:02:00.1: FW LLDP is enabled i40e 0000:02:00.1: PCI-Express: Speed 8.0GT/s Width x4 i40e 0000:02:00.1: PCI-Express bandwidth available for this device may be insufficient for optimal performance. i40e 0000:02:00.1: Please move the device to a different PCI-e link with more lanes and/or higher transfer rate. i40e 0000:02:00.1: Features: PF-id[1] VFs: 64 VSIs: 66 QP: 119 RSS FD_ATR FD_SB NTUPLE DCB VxLAN Geneve PTP VEPA i40e 0000:02:00.0 enp2s0f0: renamed from eth0 i40e 0000:02:00.1 enp2s0f1: renamed from eth1 i40e 0000:02:00.0: entering allmulti mode. This message stands out there: i40e 0000:02:00.1: PCI-Express bandwidth available for this device may be insufficient for optimal performance. i40e 0000:02:00.1: Please move the device to a different PCI-e link with more lanes and/or higher transfer rate. Does somebody has any idea ? GRUB_CMDLINE_LINUX_DEFAULT="iommu=soft" is sometimes recommended in similar cases with IO_PAGE_FAULT. Maybe I should lower the speed from 10 Gbit/s to 1 Gbit/s as a test? Thanks Stefan _______________________________________________ E1000-devel mailing list E1000-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/e1000-devel To learn more about Intel Ethernet, visit https://community.intel.com/t5/Ethernet-Products/bd-p/ethernet-products