Hi there,
I'm trying to do passthrough on another machine, since I can do the same
thing on the original, I think it would be easy, but I was wrong.
Here's the info about this server.
CPU: Intel Xeon E5 2620 v2 x 2
OS: Ubuntu 16.04 Server (Kernel ver: 4.4.0)
QEMU verison: 2.5.0 (Debian 1:2.5+dfsg-5ubuntu10.14)
Libvirt version: 1.3.1
Only the different things is this server included BMC VGA, and I install
Xfce virtual GUI + VNC.
Below is what I'm setting on Ubuntu:
1. Blacklist the nouveau or other drivers about Nvidia
Included nouveau, vga16fb, rivafb, nvidiafb, rivatv.
2. Add VFIO drivers into init list when boot .
# cat /etc/modules
vfio
vfio_pci
vfio_iommu_type1
vfio_virqfd
3. Add K420 PID & VID and open IOMMU function into GRUB
GRUB_CMDLINE_LINUX_DEFAULT="intel_iommu=on vfio-pci.ids=10de:0ff3,10de:0e1b"
4. Build VM and install Windows Server 2012 R2
5. Shutdown VM, add kvm=off, delete everything about Hyper-V function, and
passthrough K420 after installation.
But when I mount K420 into VM, VFIO shows error on kernel log.
vfio-pci 0000:06:00.0: BAR 3: can't reserve [mem 0xd0000000-0xd1ffffff
64bit pref]
And I checked /proc/iomem, it shows
80000000-dfffffff : PCI Bus 0000:00
80000000-8fffffff : PCI MMCONFIG 0000 [bus 00-ff]
80000000-8fffffff : reserved
c0000000-d1ffffff : PCI Bus 0000:06
c0000000-cfffffff : 0000:06:00.0
d0000000-d1ffffff : 0000:06:00.0
d1000000-d112ffff : vesafb
VM can boot, but show Code 43 after installed driver.
I don't know why it occupied by vesafb, since this server already have BMC
chip.
Here's the VM profile:
https://pastebin.com/PXuT4Vqg
Any ideas?
Many thanks,
Eddie.
_______________________________________________
vfio-users mailing list
vfio-users@redhat.com
https://www.redhat.com/mailman/listinfo/vfio-users
