[vfio-users] guest GPU p2p doesn't work
Alex Williamson
alex.williamson at redhat.com
Wed Aug 22 14:32:43 UTC 2018
On Wed, 22 Aug 2018 13:53:19 +0000
Zhiyong WU 吴志勇 <zhiyong.wu at bitmain.com> wrote:
> HI
>
> Today I tried to play guest GPU p2p with the following way, but failed, does anyone know the reason?
>
>
> 1. Hypervisor info
> [root at localhost ~]#./qemu-system-x86_64 --version
> QEMU emulator version 2.12.1 (v2.12.1-dirty)
> Copyright (c) 2003-2017 Fabrice Bellard and the QEMU Project developers
>
> [root at localhost ~]# nvidia-smi topo -p2p r
> GPU0 GPU1 GPU2 GPU3 GPU4 GPU5 GPU6 GPU7
> GPU0 X OK OK OK OK OK OK OK
> GPU1 OK X OK OK OK OK OK OK
> GPU2 OK OK X OK OK OK OK OK
> GPU3 OK OK OK X OK OK OK OK
> GPU4 OK OK OK OK X OK OK OK
> GPU5 OK OK OK OK OK X OK OK
> GPU6 OK OK OK OK OK OK X OK
> GPU7 OK OK OK OK OK OK OK X
>
> Legend:
>
> X = Self
> OK = Status Ok
> CNS = Chipset not supported
> GNS = GPU not supported
> TNS = Topology not supported
> NS = Not supported
> U = Unknown
> [root at localhost ~]#
>
> [root at localhost ~]# ps -ef | grep qemu
> root 2608 1 7 03:32 ?
> 00:09:42 /usr/local/qemu-2.12.1/bin/qemu-system-x86_64 -enable-kvm
> -cpu host,kvm=off -chardev
> socket,id=hmqmondev,port=55901,host=127.0.0.1,nodelay,server,nowait
> -mon chardev=hmqmondev,id=hmqmon,mode=readline -rtc
> base=utc,clock=host,driftfix=none -daemonize -nodefaults -nodefconfig
> -no-kvm-pit-reinjection -global kvm-pit.lost_tick_policy=discard
> -machine pc,accel=kvm -k en-us -smp 32 -name BarzHsu-AI -m 131072
> -boot order=cdn -device virtio-serial -usb -device usb-kbd -device
> usb-tablet -vga std -vnc :1 -device virtio-scsi-pci,id=scsi -drive
> file=/opt/cloud/workspace/disks/3691b8d4-04bd-4338-8134-67620d37bdc8,if=none,id=drive_0,cache=none,aio=native
> -device scsi-hd,drive=drive_0,bus=scsi.0,id=drive_0 -drive
> file=/opt/cloud/workspace/disks/24dc552b-8518-4334-92c8-f78c4db8f626,if=none,id=drive_1,cache=none,aio=native
> -device scsi-hd,drive=drive_1,bus=scsi.0,id=drive_1 -device
> vfio-pci,host=07:00.0,multifunction=on,addr=0x15,x-nv-gpudirect-clique=1
> -device vfio-pci,host=07:00.1 -device
Note that unless you provide an addr= for the audio function to place
it into the same slot as the GPU function, specifying multifunction=on
for the GPU function is pointless.
> vfio-pci,host=08:00.0,multifunction=on,addr=0x16,x-nv-gpudirect-clique=1
> -device vfio-pci,host=08:00.1 -device
> vfio-pci,host=04:00.0,multifunction=on,addr=0x17,x-nv-gpudirect-clique=1
> -device vfio-pci,host=04:00.1 -device
> vfio-pci,host=06:00.0,multifunction=on,addr=0x18,x-nv-gpudirect-clique=1
> -device vfio-pci,host=06:00.1 -device
> vfio-pci,host=0f:00.0,multifunction=on,addr=0x19,x-nv-gpudirect-clique=1
> -device vfio-pci,host=0f:00.1 -device
> vfio-pci,host=0e:00.0,multifunction=on,addr=0x1a,x-nv-gpudirect-clique=1
> -device vfio-pci,host=0e:00.1 -device
> vfio-pci,host=0d:00.0,multifunction=on,addr=0x1b,x-nv-gpudirect-clique=1
> -device vfio-pci,host=0d:00.1 -device
> vfio-pci,host=0c:00.0,multifunction=on,addr=0x1c,x-nv-gpudirect-clique=1
> -device vfio-pci,host=0c:00.1 -device
> ide-cd,drive=ide0-cd0,bus=ide.1,unit=1 -drive
> id=ide0-cd0,media=cdrom,if=none -netdev
> type=tap,id=vnet22-254,ifname=vnet22-254,vhost=on,vhostforce=off,script=/opt/cloud/workspace/servers/6af6cf5b-5c97-426d-92a6-972c0c40c78a/if-up-br0-vnet22-254.sh,downscript=/opt/cloud/workspace/servers/6af6cf5b-5c97-426d-92a6-972c0c40c78a/if-down-br0-vnet22-254.sh
> -device
> virtio-net-pci,netdev=vnet22-254,mac=00:22:4c:50:fe:65,addr=0xf,speed=10000
> -pidfile /opt/cloud/workspace/servers/6af6cf5b-5c97-426d-92a6-972c0c40c78a/pid
> -chardev
> socket,path=/opt/cloud/workspace/servers/6af6cf5b-5c97-426d-92a6-972c0c40c78a/qga.sock,server,nowait,id=qga0
> -device virtserialport,chardev=qga0,name=org.qemu.guest_agent.0
> -object rng-random,filename=/dev/random,id=rng0 -device
> virtio-rng-pci,rng=rng0,max-bytes=1024,period=1000
>
>
> 1. Guest info
> 2. guest at BarzHsu-AI:~$ nvidia-smi topo -m
> 3. GPU0 GPU1 GPU2 GPU3 GPU4 GPU5 GPU6 GPU7 CPU Affinity
> 4. GPU0 X PHB PHB PHB PHB PHB PHB PHB 0-31
> 5. GPU1 PHB X PHB PHB PHB PHB PHB PHB 0-31
> 6. GPU2 PHB PHB X PHB PHB PHB PHB PHB 0-31
> 7. GPU3 PHB PHB PHB X PHB PHB PHB PHB 0-31
> 8. GPU4 PHB PHB PHB PHB X PHB PHB PHB 0-31
> 9. GPU5 PHB PHB PHB PHB PHB X PHB PHB 0-31
> 10. GPU6 PHB PHB PHB PHB PHB PHB X PHB 0-31
> 11. GPU7 PHB PHB PHB PHB PHB PHB PHB X 0-31
> 12.
> 13. Legend:
> 14.
> 15. X = Self
> 16. SYS = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)
> 17. NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node
> 18. PHB = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)
> 19. PXB = Connection traversing multiple PCIe switches (without traversing the PCIe Host Bridge)
> 20. PIX = Connection traversing a single PCIe switch
> 21. NV# = Connection traversing a bonded set of # NVLinks
> 22.
> guest at BarzHsu-AI:~$ nvidia-smi topo -p2p r
>
> 1. GPU0 GPU1 GPU2 GPU3 GPU4 GPU5 GPU6 GPU7
> 2. GPU0 X CNS CNS CNS CNS CNS CNS CNS
> 3. GPU1 CNS X CNS CNS CNS CNS CNS CNS
> 4. GPU2 CNS CNS X CNS CNS CNS CNS CNS
> 5. GPU3 CNS CNS CNS X CNS CNS CNS CNS
> 6. GPU4 CNS CNS CNS CNS X CNS CNS CNS
> 7. GPU5 CNS CNS CNS CNS CNS X CNS CNS
> 8. GPU6 CNS CNS CNS CNS CNS CNS X CNS
> 9. GPU7 CNS CNS CNS CNS CNS CNS CNS X
> 10.
> 11. Legend:
> 12.
> 13. X = Self
> 14. OK = Status Ok
> 15. CNS = Chipset not supported
> 16. GNS = GPU not supported
> 17. TNS = Topology not supported
> 18. NS = Not supported
> 19. U = Unknown
Does "Chipset not supported" suggest NVIDIA has disabled GPU Direct for
i440fx regardless of providing the clique information? Maybe try
QEMU's q35 chipset, perhaps even in a more typical PCIe configuration
with GPUs downstream of root ports. Perhaps NVIDIA no longer honors
their own specification. Dunno. Thanks,
Alex
More information about the vfio-users
mailing list