Is it possible to enable virtio pci drivers on VM hosted by TX2?

Autonomous Machines Jetson & Embedded Systems Jetson TX2
1

Hi everyone,

Thanks to gattis and marcus_c, i was able to launch a VM with KVM/Qemu on tx2 eventually.

To go further in my current work, i need to enable virtio pci driver on the VM.

Here are things what i had done:

  1. host kernel config [kernel version:4.4.38]
#nvidia@tegra-ubuntu:/boot$ zgrep VIRTIO /proc/config.gz
CONFIG_VIRTIO_BLK=y
# CONFIG_SCSI_VIRTIO is not set
# CONFIG_VIRTIO_NET is not set
# CONFIG_VIRTIO_CONSOLE is not set
CONFIG_TRUSTY_VIRTIO=y
CONFIG_TRUSTY_VIRTIO_IPC=y
CONFIG_VIRTIO=y
CONFIG_VIRTIO_PCI=y
CONFIG_VIRTIO_PCI_LEGACY=y
CONFIG_VIRTIO_BALLOON=y
CONFIG_VIRTIO_INPUT=y
CONFIG_VIRTIO_MMIO=y
# CONFIG_VIRTIO_MMIO_CMDLINE_DEVICES is not set
  1. guest kerenl config [kernel version:3.19.0-28-generic]
root@localhost:~$ cat /boot/config-3.19.0-28-generic | grep -i "virtio"
CONFIG_NET_9P_VIRTIO=m
CONFIG_VIRTIO_BLK=y
CONFIG_SCSI_VIRTIO=m
CONFIG_VIRTIO_NET=y
CONFIG_CAIF_VIRTIO=m
CONFIG_VIRTIO_CONSOLE=y
CONFIG_HW_RANDOM_VIRTIO=m
CONFIG_VIRTIO=y
# Virtio drivers
CONFIG_VIRTIO_PCI=y
CONFIG_VIRTIO_BALLOON=y
CONFIG_VIRTIO_MMIO=y
CONFIG_VIRTIO_MMIO_CMDLINE_DEVICES=y
  1. Run Qemu as follow
taskset 0x39 $QEMU  -machine virt  -cpu cortex-a57 \                                                                              
               -smp 4 -m 4096 \                                                                                                      
               -enable-kvm \                                                                                                         
               -device virtio-scsi-device,id=scsi \                                                                                  
               -device virtio-scsi-pci,id=pci \                                                                                      
               -kernel vmlinuz-3.19.0-28-generic \                                                                                   
               -initrd initrd.img-3.19.0-28-generic \                                                                                                                                                                          
               -device scsi-hd,drive=coreimg \                                                                                       
               -drive file="$DISK_FILE",id=coreimg,cache=unsafe,if=none \                                                                                                                                                    
               -netdev user,id=unet -device virtio-net-device,netdev=unet  \                                                         
               --append "console=ttyAMA0 root=/dev/sda" \                                                                            
               --redir tcp:2228::22 \                                                                                                
               -vnc :2 \                                                                                                             
               -device VGA -device ne2k_pci

However, run sudo lspci -nvv on guest show nothing!?
I had try similar setup on X86 environment, sudo lspci -nvv will output

... ...
Kernel driver in use: virtio-pci
...

Questions:

  1. Does it related to dtb files? cause in order to boot tx2 in hyp mode, i had change some .dtsi file.
  2. Possible some loss of kernel config options?
  3. Anything else I can try?

Thanks for your time ^0^

2

We are not yet providing VM support. Maybe other user can share their experience here.

3

My modification can let TX2’s a57 cores boot into Hyp mode, work fine together with KVM/QEMU.
Is TX2 not compatible with VIRTIO framework?

4

Hi jordan.huang,

I have virtio+9p working fine in my KVM guest, so it is definitely possible to use.

Since you built virtio as a module, did you remember to load it in the guest?

I’m using Linux 4.14.15 for the guest. Here’s my guest kernel configuration: Kernel configuration for big endian aarch64 KVM guest · GitHub

Note that my guest is big endian, you can change that if you want to.

This is my qemu invocation line:

taskset 0x39 qemu-system-aarch64 -machine virt -cpu host -enable-kvm -nographic -smp 1 -m 2048 \
  -virtfs local,path=/usr/aarch64_be-unknown-linux-gnu,mount_tag=root9p,security_model=passthrough,id=root9p \
  -virtfs local,path=/home,mount_tag=home9p,security_model=passthrough,id=home9p \
  -kernel "$KERNEL" -append "$CMDLINE"

From my guest I can do

sh-4.4# lspci
00:00.0 Host bridge: Red Hat, Inc. QEMU PCIe Host bridge
00:01.0 Ethernet controller: Red Hat, Inc. Virtio network device
00:02.0 Unclassified device [0002]: Red Hat, Inc. Virtio filesystem
00:03.0 Unclassified device [0002]: Red Hat, Inc. Virtio filesystem
sh-4.4#
5

Hi marcus_c,

Thank you for your continued support, it really help me a lot. :)

By comparing my kernel config with yours, i found something difference related to “IOMMU”, and directly update the guest kernel to ver.4.2.0-18 makes it work!

6

Hi all,

I successfully create virtual machines using KVM on Jetson TX2. If you are interested, please see my Github repository for detailed information.

Baoqian Wang

7

Hi Baoqian,

Even the VM is not we suggested, but thanks anyway for the sharing.

8

How to confim that GPU passthough worked for the VM after VM installation is succesfull?

9

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.