[vfio-users] vga passthrough Zotac GTX 1060 3G

Alex Williamson alex.williamson at redhat.com
Mon Mar 13 18:22:57 UTC 2017


On Fri, 10 Mar 2017 22:33:15 +0100
Jiri Novak <jiri.novak at ghorland.net> wrote:

> Hi,
> 
> previuosly I've successfully passed radeon cards on the same box. now
> I'm trying to pass Zotac GTX 1060 3G (both to new or the old VMs) and
> can't get past:
> Running option rom at c000:0003 - if I don't specify romfile
> Booting from 0000:7c00 - if I specify romfile, the VM doesn't boot though.
> no output at all on serial console if I try to boot with -bios
> /usr/share/qemu/OVMF.fd
> 
> logs with and without romfile in attachements, those would be long
> 
> any ideas what to try next?

Is the host platform Intel with IGD graphics enabled?  If so then VGA
accesses are probably accessing IGD instead of the assigned GPU, which
can cause such problems.  You're better off getting OVMF to work
properly (-bios is not the right way to invoke OVMF).

> the qemu commandline (if you have any other hints regarding better
> options, I'd appreciate it. especially if I can better map what cpus are
> used for which vcpu):

Even better, use libvirt so you can pin individual vCPUs.

> 
> LC_ALL=C
> PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
> QEMU_AUDIO_DRV=pa /usr/bin/taskset -c 4-7,20-23
> /usr/bin/qemu-system-x86_64 \
>     -name 12-wingame \
>     -machine pc-q35-2.4,accel=kvm,usb=off \
>     -cpu
> SandyBridge,+invtsc,+osxsave,+pcid,+pdcm,+xtpr,+tm2,+est,+smx,+vmx,+ds_cpl,+monitor,+dtes64,+pbe,+tm,+ht,+ss,+acpi,+ds,+vme,hv_time,hv_relaxed,hv_vapic,hv_spinlocks=0x1fff,kvm=off
> \
>     -object
> memory-backend-file,id=mem0,size=16G,mem-path=/dev/hugepages,share=off \
>     -numa node,nodeid=0,memdev=mem0 \
>     -m 16G \
>     -realtime mlock=off \
>     -smp sockets=1,cores=4,threads=2 \
>     -uuid aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaa \
>     -nographic -no-user-config -nodefaults -no-hpet \
>     -rtc base=localtime,driftfix=slew \
>     -global kvm-pit.lost_tick_policy=discard \
>     -boot d \
>     -device
> ioh3420,bus=pcie.0,addr=1c.0,multifunction=on,port=1,chassis=1,id=root.1 \
>     -device pci-bridge,chassis_nr=2,id=pci.2,bus=pcie.0,addr=0x3 \
>     -device pci-bridge,chassis_nr=3,id=pci.3,bus=pcie.0,addr=0x4 \
>     -device pci-bridge,chassis_nr=4,id=pci.4,bus=pcie.0,addr=0x5 \
>     -device pci-bridge,chassis_nr=5,id=pci.5,bus=pcie.0,addr=0x6 \

What's going on here?  Why attach so many pci-bridges?  These aren't
PCIe root ports.

>     -device piix3-usb-uhci,id=usb,bus=pcie.0,addr=0x7 \
>     -drive
> file=/dev/Z-ssd/qemu/12-wingame,if=none,id=drive-virtio-disk0,format=raw,cache=none,aio=native
> \
>     -drive
> file=/dev/Z-ssd/data/games-blizzard,if=none,id=drive-virtio-disk1,format=raw,cache=none,aio=native
> \
>     -device virtio-scsi-pci,bus=pci.2,addr=0x1 \
>     -device scsi-hd,drive=drive-virtio-disk0 \
>     -device scsi-hd,drive=drive-virtio-disk1 \
>     -netdev tap,id=netdev0,ifname=V0300t12,script=no,downscript=no \
>     -device
> virtio-net-pci,netdev=netdev0,id=net0,mac=42:42:42:00:00:0c,bus=pci.2,addr=0x2
> \
>     -device vfio-pci,host=07:00.0,id=hostdev2,bus=pci.4,addr=0x1 \
>     -device vfio-pci,host=00:1a.0,id=hostdev3,bus=pci.2,addr=0x4 \
>     -device vfio-pci,host=00:1d.0,id=hostdev5,bus=pci.2,addr=0x6 \
>     -device
> vfio-pci,host=02:00.0,id=hostdev6,bus=pci.5,multifunction=on,addr=0x1,x-vga=on
> \
>     -device vfio-pci,host=02:00.1,id=hostdev7,bus=pci.5,addr=0x1.0x1 \
>     -device virtio-balloon-pci,id=balloon0,bus=pci.2,addr=0x7 \

Strange layout.  What benefit are you trying to achieve here?

>     -vga none \
>     -soundhw hda \
>     -chardev stdio,id=seabios \
>     -device isa-debugcon,iobase=0x402,chardev=seabios \
>     -msg timestamp=on \
>         -drive
> file=/mnt/X-gzfs/backups/pools/C-nas/data/sysiso/win10n_1607.iso,index=0,media=cdrom
> \
>         -drive
> file=/mnt/X-gzfs/backups/pools/C-nas/data/sysiso/virtio-win-0.1.118.iso,index=1,media=cdrom
> \
>     >>/var/log/kvm/12-wingame.stdout 2>>/var/log/kvm/12-wingame.stderr &  
> 
> Thanks in advance
> Gh.




More information about the vfio-users mailing list