Pci passthrough with qemu

Http://pic.dhe.ibm.com/infocenter/lnxinfo/v3r0m0/index.jsp? Topic=%2fliaat%2fliaatbppassthrougtask.htm

Http://www.linux-kvm.org/page/How_to_assign_devices_with_VT-d_in_KVM

1. Check whether the hardware supports iommu.

This is a hardware function. It is called VT-D in Intel architecture.

It can be seen from the BIOS that the positions of different versions are different.

My website is under security-> virtualization-> VT-D.

2. Configure the kernel and support iommu

There is such an option on Intel machines,

Intel_iommu

│ Location: │
│-> Device Drivers │
│ (2)-> IOMMU Hardware Support (IOMMU_SUPPORT [= y])

3. Restart and check whether the service is enabled.

On intel, run

Dmesg | grep-e DMAR-e IOMMU

4. Unbind Device on host

Load the pci_stub module, modprobe pci_stub

Select a device, lspci-s. 0-n,. 0 0280: 8086: 0085 (rev 34)

Bind/unbind

• Echo "8086 0085">/sys/bus/PCI/Drivers/PCI-stub/new_id
• Echo response :0:00. 0>/sys/bus/PCI/devices/response :0:00. 0/driver/unbind
• Echo restart :0:00. 0>/sys/bus/PCI/Drivers/PCI-stub/Bind

5. Send the device to guest

./I386-softmmu/qemu-system-i386-enable-kvm \

-Drive file =.../../kvm/ubuntu. qcow2-boot dc-m 512 \

-Monitor stdio \

-Netdev user, id = usernet-device rtl8139, netdev = usernet \

-Device pci-assign, host =. 0

But even if it fails, an error is returned...

6. Error Analysis and Solution

According to the above method, I did not succeed. I read dmesg, and there is such a sentence.

Kvm_iommu_map_guest: No interrupt remapping support, disallowing device assignment. Re-enble with "allow_unsafe_assigned_interrupts = 1" module option.

After a slight investigation, we found that it was related to this option, config IRQ_REMAP. Unfortunately, I cannot enable it because it must be done on x86_64.

No way, I have to add this parameter when loading the kvm module.

After this parameter is added, it is enough. Like.

Next I used vfio for passthrough. The steps are slightly different..

1. Install the relevant kernel module

Sudo modprobe vfio

Sudo modprobe vfio-pci

In this way, the/dev/vfio device is generated.

2. Add the device to a group according to the document requirements.

Echo 0000: 0d: 00.0>/sys/bus/pci/devices/0000: 0d: 00.0/driver/unbind

Echo 1180 e823>/sys/bus/pci/drivers/vfio-pci/new_id

Echo response :0:00. 0>/sys/bus/pci/devices/response :0:00. 0/driver/unbind

Echo 8086 0085>/sys/bus/pci/drivers/vfio-pci/new_id

In this way,/dev/vfio/6 is generated (the number is determined by the iommu_group number)

3. Start guest

.. /Git/qemu-latest/i386-softmmu/qemu-system-i386-enable-kvm-drive file = ubuntu. qcow2-boot dc-m 512-monitor stdio-netdev user, id = usernet-device vfio-pci, host =. 0

If it is on the power machine, it is written

-Device "spapr-pci-vfio-host-bridge, id = CXGB3,Iommu = 4, Index = 6"

4. fix

By default, this will cause an error. Of course, this is on my machine.

Vfio_iommu_type1_open: No interrupt remapping support. Use the module param "allow_unsafe_interrupts" to enable vfio iommu support on this platform

This means that my machine does not support REMAP, so I just set this parameter to 1.