Configuration:
Host: CentOS 6.4
Mellanox Card: ConnectX-3 EN
Mellanox Driver: MLNX_OFED_LINUX-2.0-3.0.0-rhel6.4-x86_64 with SRIOV enabled
Virtualization: KVM
Problem:
Have up to 7 virtual functions enabled and running smoothly for 14 days. Suddenly some of the virtual functions do not come up on boot time inside the guests(ifconfig does not show the interface, but ifconfig -a shows the interface). It only happens with only certain virtual functions.Also rebooted the guests, reinstalled the guests, problem still persists.
Looking at the host's syslog got this message printed only when a virtual machine with a faulty virtual functions were booted:
May 13 11:12:22 localhost kernel: mlx4_core 0000:09:00.0: vhcr command ALLOC_RES (0xf00) slave:7 in_param 0x0 in_mod=0x1, op_mod=0x1 failed with error:0, status -22
Workaround:
In host I removed all the guests and reloaded the mellanox drivers(mlx4_en and mlx4_core), now everything works as expected.
Question:
Is this a bug in the driver? Is this got fixed in the latest driver?
Thanks, help is appreciated!