[s3-bug-review] question about virtio related options: ats=on and iommu=on

[Han Han]

On Mon, Jan 9, 2023 at 12:01 PM Yalan Zhang <yalzhang at redhat.com> wrote:

> Hi,
>
> I have a question about the virtio related options, could someone please
> help to confirm?
> In my understanding, the ats='on' should depend on the "iommu=on", but I'm
> not sure about it.
> Please check the details below.
> Thank you!
>
> Details:
> 1. set vm with intel iommu device, and enable iommu for virtio
> filesystem  device:
> # virsh dumpxml rhel  --xpath //iommu
> <iommu model="intel">
>   <driver intremap="on" caching_mode="on" iotlb="on"/>
> </iommu>
>
> Set iommu="on" ats="on" for virtio filesystem device:
> # virsh dumpxml rhel --xpath //filesystem
> <filesystem type="mount" accessmode="passthrough">
>   <driver type="virtiofs" queue="512" iommu="on" ats="on"/>
>   <binary path="/usr/libexec/virtiofsd" xattr="on">
>     <cache mode="none"/>
>   </binary>
>   <source dir="/path1"/>
>   <target dir="mount_tag1"/>
>   <address type="pci" domain="0x0000" bus="0x04" slot="0x00"
> function="0x0"/>
> </filesystem>
>
> 2. failed to start the vm since "iommu_platform=true is not supported by
> the device"
> # virsh start rhel
> error: Failed to start domain 'rhel'
> error: internal error: qemu unexpectedly closed the monitor:
> 2023-01-09T03:48:03.198629Z qemu-kvm: -device
> {"driver":"vhost-user-fs-pci","iommu_platform":true,"ats":true,"id":"fs0","chardev":"chr-vu-fs0","queue-size":512,"tag":"mount_tag1","bus":"pci.4","addr":"0x0"}:
> iommu_platform=true is not supported by the device
>
> The error msg comes from the qemu commit
https://gitlab.com/qemu-project/qemu/-/commit/e65902a913bf31ba79a83a3bd3621108b85cf645
See the commit msg for more info

> 3. update the xml to be with only ats="on", vm can start successfully,
> which is not expected.
> # virsh dumpxml rhel --xpath //filesystem
> <filesystem type="mount" accessmode="passthrough">
>   <driver type="virtiofs" queue="512" ats="on"/>
>   <binary path="/usr/libexec/virtiofsd" xattr="on">
>     <cache mode="none"/>
>   </binary>
>   <source dir="/path1"/>
>   <target dir="mount_tag1"/>
>   <alias name="fs0"/>
>   <address type="pci" domain="0x0000" bus="0x04" slot="0x00"
> function="0x0"/>
> </filesystem>
>
> the qemu cmd line:
> -chardev
> socket,id=chr-vu-fs0,path=/var/lib/libvirt/qemu/domain-5-rhel/fs0-fs.sock \
> -device
> '{"driver":"vhost-user-fs-pci","ats":true,"id":"fs0","chardev":"chr-vu-fs0","queue-size":512,"tag":"mount_tag1","bus":"pci.4","addr":"0x0"}'
>
> Refer to libvirt.org:
> QEMU's virtio devices have some attributes related to the virtio transport
> under the driver element: The iommu attribute enables the use of emulated
> IOMMU by the device.
> The attribute ats controls the Address Translation Service support for
> PCIe devices. This is needed to make use of IOTLB support (see IOMMU
> devices). Possible values are on or off.
>
> In my understanding, the ats='on' should depend on the iommu='on', so if
> the device does not support iommu, it should not support ats, eigher.
> I'm not sure if the understanding is correct.
>
Not really. ats means Address Translation Services
<https://www.intel.com/content/www/us/en/docs/programmable/683686/20-4/address-translation-services-ats.html>
a capability for PCIe device.
iommu means device can be used on IOMMU platform
<https://github.com/qemu/qemu/blob/master/qapi/virtio.json#L315>

>
> Could someone please help to confirm it?
> Thank you!
>
>
> Yalan
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/libvirt-users/attachments/20230109/a04d30f8/attachment.htm>