Debugging a General Protection Fault in the i40e driver when adding a VLAN to a VF, I added a ton of dev_info() calls to print out data and determine some of the code flow. I am using the Intel SR-IOV CNI ( https://github.com/intel/sriov-cni/ and https://github.com/intel/sriov-network-device-plugin) to created and delete pods with SR-IOV VFs attached to containers.
I think there is an issue on how the VLAN, the QoS and the vlanprio (vlanprio = vlan_id | (qos << I40E_VLAN_PRIORITY_SHIFT) are handled in the i40e driver. When a VLAN is added, a call is made to i40e_ndo_set_vf_port_vlan() in i40e_virtchnl_pf.c. Input to the function includes a VLAN (lets say 200) and a QoS value (lets say 1). The VLAN is at some point added the mac_filter_hash. When the function exits, the value in the mac_filter_hash ends up as 8392 (which is the vlanprio). Here is a dev_err() message I added at the beginning and end of i40e_ndo_set_vf_port_vlan(), just before returning. The last log is a loop through the mac_filter_hash. The VLAN in the hash is 8392. [ +0.032191] i40e 0000:01:00.1: BILLY: i40e_ndo_set_vf_port_vlan() ENTER - vf_id=0 VLAN=200 qos=1 vlanprio=8392 (0x20c8) : [ +0.000921] i40e 0000:01:00.1: BILLY - i40e_ndo_set_vf_port_vlan() - EXIT: vid 200: f=0000000073750ac4 VLANID=8392 State=2 next=00000000ce90cad7 pprev=00000000949c824a prev=0000000073750ac4 The issue is throughout the code, the f->vlan (which is the VLAN in mac_filter_hash) is treated as VLAN Id. So if it is vlanprio, range checks fail because the priority is not masked off. For example: if (f->vlan < 0 || f->vlan > I40E_MAX_VLANID) So when time to delete the VLAN, by passing VLAN 0 to i40e_ndo_set_vf_port_vlan(), it is not deleted. Same log on a delete: [Sep17 10:19] i40e 0000:01:00.1: BILLY: ENTER - vf_id=0 VLAN=0 qos=0 vlanprio=0 (0x0000) : [ +0.000384] i40e 0000:01:00.1: BILLY - i40e_ndo_set_vf_port_vlan - EXIT: vid 0: f=000000001221c957 VLANID=0 State=2 next=0000000073750ac4 pprev=00000000949c824a prev=000000001221c957 [ +0.000018] i40e 0000:01:00.1: BILLY - i40e_ndo_set_vf_port_vlan - EXIT: vid 0: f=0000000073750ac4 VLANID=8392 State=2 next=00000000ce90cad7 pprev=000000001221c957 prev=0000000073750ac4 System Details: Running Fedora 29 on baremetal (DELL R630), kernel 5.2.11-100.fc29.x86_64 $ ethtool -i eno1 driver: i40e version: 2.9.21 firmware-version: 6.80 0x80003d71 18.8.9 expansion-rom-version: bus-info: 0000:01:00.0 supports-statistics: yes supports-test: yes supports-eeprom-access: yes supports-register-dump: yes supports-priv-flags: yes Let me know what other details I can provide. Thanks, Billy McFall _______________________________________________ E1000-devel mailing list E1000-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/e1000-devel To learn more about Intel® Ethernet, visit http://communities.intel.com/community/wired
