I recently did setup a new XCP-NG host, which has an Intel X520-DA2 10GBit card in it.
Transceivers are from Finisar.
The ixgbe driver in its current configuration does not recognzie these non-Intel SFPs:
[00:56 vmcluster04 ~]# dmesg | grep SFP
[ 11.367930] ixgbe 0000:03:00.0: failed to load because an unsupported SFP+ or QSFP module type was detected.
[ 11.400061] ixgbe 0000:03:00.1: failed to load because an unsupported SFP+ or QSFP module type was detected.
[00:56 vmcluster04 ~]# #
I know, that I can allow non-Intel SFPs by adding a kernel option “allow_unsupported_sfp=1”, however: I did not figure out on how to do that on the XCP-NG environment.
Trying to remove the module and insert it again with the parameter passed to the command failed also:
[00:56 vmcluster04 ~]# rmmod ixgbe
[01:00 vmcluster04 ~]# insmod /usr/lib/modules/4.19.0+1/kernel/drivers/net/ethernet/intel/ixgbe/ixgbe.ko allow_unsupported_sfp=1
insmod: ERROR: could not insert module /usr/lib/modules/4.19.0+1/kernel/drivers/net/ethernet/intel/ixgbe/ixgbe.ko: Unknown symbol in module
[01:00 vmcluster04 ~]#
I tried to add this to /etc/modprobe.d/ixgbe-options.conf:
options ixgbe allow_unsupported_sfp=1
and after doing a:
rmmod ixgbe
modprobe ixgbe
I could find this in the dmesg log:
[14:44 vmcluster04 ~]# dmesg
[ 641.481979] Intel(R) 10GbE PCI Express Linux Network Driver - version 5.5.2
[ 641.481981] Copyright(c) 1999 - 2018 Intel Corporation.
[ 641.482211] xen: registering gsi 30 triggering 0 polarity 1
[ 641.482228] Already setup the GSI :30
[ 641.500233] ixgbe: 0000:03:00.0: ixgbe_check_options: FCoE Offload feature enabled
[ 641.500235] ixgbe: allow_unsupported_sfp Enabled
[ 641.511343] ixgbe 0000:03:00.0: failed to load because an unsupported SFP+ or QSFP module type was detected.
[ 641.511361] ixgbe 0000:03:00.0: Reload the driver after installing a supported module.
[ 641.511760] xen: registering gsi 37 triggering 0 polarity 1
[ 641.511776] Already setup the GSI :37
[ 641.532226] ixgbe: 0000:03:00.1: ixgbe_check_options: FCoE Offload feature enabled
[ 641.532227] ixgbe: allow_unsupported_sfp Enabled
[ 641.543354] ixgbe 0000:03:00.1 0000:03:00.1 (uninitialized): WARNING: Intel (R) Network Connections are quality tested using Intel (R) Ethernet Optics. Using untested modules is not support ed and may cause unstable operation or damage to the module or the adapter. Intel Corporation is not responsible for any harm caused by using untested modules.
[ 641.681819] ixgbe 0000:03:00.1: Multiqueue Enabled: Rx Queue count = 16, Tx Queue count = 16 XDP Queue count = 0
[ 641.684051] ixgbe 0000:03:00.1: 32.000 Gb/s available PCIe bandwidth (5 GT/s x8 link)
[ 641.684394] ixgbe 0000:03:00.1 eth2: MAC: 2, PHY: 17, SFP+: 6, PBA No: G28774-000
[ 641.684397] ixgbe 0000:03:00.1: a0:36:9f:27:9f:0a
[ 641.684414] ixgbe 0000:03:00.1 eth2: Enabled Features: RxQ: 16 TxQ: 16 FdirHash
[ 641.686499] ixgbe 0000:03:00.1 eth2: Intel(R) 10 Gigabit Network Connection
[ 641.731152] ixgbe 0000:03:00.1 side-4034-eth2: renamed from eth2
[14:44 vmcluster04 ~]#
So, the option itself results at least in a change of behaviour, however, only one of the two 10G ports seems to be detected on PCI slot 0000:03:00.1 - the other one at 0000:03:00.0 is allegedly still equipped with an unsupported SFP module - but it’s the same one as in the other port.
Also, according to the dmesg log, one of these interfaces has been added as eth2, but when trying to acquire this interface:
