Re: Host device assignment driver name vfio/ kvm

[Moshe Levi <moshele@xxxxxxxxxxxx>]

 

 

From: sendmail [mailto:justsendmailnothingelse@xxxxxxxxx] On Behalf Of Laine Stump
Sent: Sunday, April 03, 2016 7:14 PM
To: Libvirt <libvir-list@xxxxxxxxxx>
Cc: Moshe Levi <moshele@xxxxxxxxxxxx>; Alex Williamson <Alex.Williamson@xxxxxxxxxx>
Subject: Re: [libvirt] Host device assignment driver name vfio/ kvm

 

(please configure your email client to properly quote earlier messages, and reply inline rather as a top-post; it makes it easier for others to determine the context of replies.)

On 04/03/2016 05:00 AM, Moshe Levi wrote:

Hi,

 

Ok, so here are the errors I get when using vfio/( kvm is working fine)

 

This is the Error [1] I am getting on resume when using vfio. And [2] is the ls output on the /dev/vfio/

Please note that I am trying to attach pci which is VF

My setup is:

Fedora 21

Libvirt 1.3.0

Qemu 2.5

Kernel  3.17.4-301.fc21.x86_64

OpenStack – master.

 

Let me know if you need more information.

[1] –

root@r-dcs78:/var/log/libvirt/qemu# cat ./instance-0000000e.log

2016-03-31 15:31:09.895+0000: starting up libvirt version: 1.3.0, package: 1.fc21 (Unknown, 2016-02-25-10:56:09, r-ufm231.mtr.labs.mlnx), qemu version: 2.5.0 (qemu-2.5.0-11.fc21)

LC_ALL=C PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin QEMU_AUDIO_DRV=none /usr/bin/qemu-kvm -name instance-0000000e -S -machine pc-i440fx-2.5,accel=kvm,usb=off -m 2048 -realtime mlock=off -smp 1,sockets=1,cores=1,threads=1 -uuid a809a3b5-ba75-41e6-889f-ffeebecfe54e -smbios 'type=1,manufacturer=OpenStack Foundation,product=OpenStack Nova,version=13.0.0,serial=533a28c3-7519-4e6d-8cdb-4b1f72649e71,uuid=a809a3b5-ba75-41e6-889f-ffeebecfe54e,family=Virtual Machine' -no-user-config -nodefaults -chardev socket,id=charmonitor,path=/var/lib/libvirt/qemu/domain-instance-0000000e/monitor.sock,server,nowait -mon chardev=charmonitor,id=monitor,mode=control -rtc base=utc,driftfix=slew -global kvm-pit.lost_tick_policy=discard -no-hpet -no-shutdown -boot strict=on -device piix3-usb-uhci,id=usb,bus=pci.0,addr=0x1.0x2 -drive file=/opt/stack/data/nova/instances/a809a3b5-ba75-41e6-889f-ffeebecfe54e/disk,if=none,id=drive-virtio-disk0,format=qcow2,cache=none -device virtio-blk-pci,scsi=off,bus=pci.0,addr=0x3,drive=drive-virtio-disk0,id=virtio-disk0,bootindex=1 -drive file=/opt/stack/data/nova/instances/a809a3b5-ba75-41e6-889f-ffeebecfe54e/disk.config,if=none,id=drive-ide0-1-1,readonly=on,format=raw,cache=none -device ide-cd,bus=ide.1,unit=1,drive=drive-ide0-1-1,id=ide0-1-1 -chardev file,id=charserial0,path=/opt/stack/data/nova/instances/a809a3b5-ba75-41e6-889f-ffeebecfe54e/console.log -device isa-serial,chardev=charserial0,id=serial0 -chardev pty,id=charserial1 -device isa-serial,chardev=charserial1,id=serial1 -vnc 127.0.0.1:0 -k en-us -device cirrus-vga,id=video0,bus=pci.0,addr=0x2 -device vfio-pci,host=04:00.7,id=hostdev0,bus=pci.0,addr=0x4 -device virtio-balloon-pci,id=balloon0,bus=pci.0,addr=0x5 -msg timestamp=on

char device redirected to /dev/pts/29 (label charserial1)

2016-03-31T15:31:46.775166Z qemu-system-x86_64: terminating on signal 15 from pid 6866

2016-03-31 15:32:07.547+0000: starting up libvirt version: 1.3.0, package: 1.fc21 (Unknown, 2016-02-25-10:56:09, r-ufm231.mtr.labs.mlnx), qemu version: 2.5.0 (qemu-2.5.0-11.fc21)

LC_ALL=C PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin QEMU_AUDIO_DRV=none /usr/bin/qemu-kvm -name instance-0000000e -S -machine pc-i440fx-2.5,accel=kvm,usb=off -m 2048 -realtime mlock=off -smp 1,sockets=1,cores=1,threads=1 -uuid a809a3b5-ba75-41e6-889f-ffeebecfe54e -smbios 'type=1,manufacturer=OpenStack Foundation,product=OpenStack Nova,version=13.0.0,serial=533a28c3-7519-4e6d-8cdb-4b1f72649e71,uuid=a809a3b5-ba75-41e6-889f-ffeebecfe54e,family=Virtual Machine' -no-user-config -nodefaults -chardev socket,id=charmonitor,path=/var/lib/libvirt/qemu/domain-instance-0000000e/monitor.sock,server,nowait -mon chardev=charmonitor,id=monitor,mode=control -rtc base=utc,driftfix=slew -global kvm-pit.lost_tick_policy=discard -no-hpet -no-shutdown -boot strict=on -device piix3-usb-uhci,id=usb,bus=pci.0,addr=0x1.0x2 -drive file=/opt/stack/data/nova/instances/a809a3b5-ba75-41e6-889f-ffeebecfe54e/disk,if=none,id=drive-virtio-disk0,format=qcow2,cache=none -device virtio-blk-pci,scsi=off,bus=pci.0,addr=0x3,drive=drive-virtio-disk0,id=virtio-disk0,bootindex=1 -drive file=/opt/stack/data/nova/instances/a809a3b5-ba75-41e6-889f-ffeebecfe54e/disk.config,if=none,id=drive-ide0-1-1,readonly=on,format=raw,cache=none -device ide-cd,bus=ide.1,unit=1,drive=drive-ide0-1-1,id=ide0-1-1 -chardev file,id=charserial0,path=/opt/stack/data/nova/instances/a809a3b5-ba75-41e6-889f-ffeebecfe54e/console.log -device isa-serial,chardev=charserial0,id=serial0 -chardev pty,id=charserial1 -device isa-serial,chardev=charserial1,id=serial1 -vnc 127.0.0.1:0 -k en-us -device cirrus-vga,id=video0,bus=pci.0,addr=0x2 -incoming defer -device virtio-balloon-pci,id=balloon0,bus=pci.0,addr=0x5 -msg timestamp=on

char device redirected to /dev/pts/29 (label charserial1)

vfio: error opening /dev/vfio/57: Permission denied

vfio: failed to get group 57

 

It's very strange that this error follows a commandline that doesn't have any "-device vfio-pci" in it, and the previous commandline (which *did* have a "-device vfio-pci") had no error message. Is the device being hotplugged? If not, why isn't it in the commandline? Can you include the XML for the device in your next reply? (Since you're assigning a network device, I'm hoping that you're using <interface type='hostdev'> rather than <hostdev> - the former allows you to set the VF's MAC address / vlan tag after detaching it from the host driver and before assigning it to the guest)

 

I upgrade my OS to Fedora 23 with kernel-4.2.3-300.fc23.x86_64 and it working now, so it maybe a kernel issue.

In my case I am using hostdev because the VF is  InfiniBand Interface, but I had the same problems with <interface type=’hostdev’> on other setup (I attached the vm xml to the mail)

Alex: what could be the reasons for permission-denied when opening the node for the iommu group? Could this be an issue with other devices being in the same iommu group? (I would hope this isn't the case, since the device in question is apparently an SRIOV VF, so it should have its own iommu group).

I'll be offline for about 6 days starting Monday AM, so may not be able to participate in this discussion much more. Good luck!

2016-03-31T16:05:54.975312Z qemu-system-x86_64: terminating on signal 15 from pid 6866

 

[2] root@r-dcs78:/var/log/libvirt/qemu# ls -l /dev/vfio/

total 0

crw-rw-rw- 1 root root 10, 196 Mar 29 11:11 vfio

 

From: sendmail [mailto:justsendmailnothingelse@xxxxxxxxx] On Behalf Of Laine Stump
Sent: Thursday, March 31, 2016 7:07 PM
To: Libvirt <libvir-list@xxxxxxxxxx>
Cc: Moshe Levi <moshele@xxxxxxxxxxxx>
Subject: Re: [libvirt] Host device assignment driver name vfio/ kvm

 

On 03/31/2016 11:58 AM, Moshe Levi wrote:

Thanks Laine,

 

Adding  the driver_name o the config did the trick,  thanks J

 

Regarding the vifo error it seem that openstack does roleback if operation failed so that why you see the virDomainAttachDeviceFlags

Anyhow I found that the in qemu 2.1 suspend is not working (I got error [1] )  so I upgrade to qemu 2.5 and then suspend work but it failed on resume.

 

So Just to clarify vfio is not supporting “attach device” right?  is qemu going to support it?

Not at all. attach device works just fine with vfio (that's how hotplug is done). I've just been using it myself in some testing. (note that vfio has been supported since kernel 3.6 and libvirt 1.0.5, i.e. a "very long time")

Now I wonder if I need to hardcode in the hostdev config to be with driver_name=kvm…

I certainly hope not. You would be viciously attacked by the keeper of vfio for your transgression! :-)

Seriously, there should be no reason whatsoever to force driver name='kvm'. This was made configurable *only* as a fallback in case errors were encountered with vfio during its early days. As a matter of fact, I'm pretty sure that RHEL7 has legacy kvm device assignment completely disabled. If you are needing to force legacy kvm device assignment to make your setup work, then there is a bug somewhere that needs to be investigated and squashed.

 

 

[1] -ERROR:qom/object.c:725:object_unref: assertion failed: (obj->ref > 0)

 

 

 

 

From: sendmail [mailto:justsendmailnothingelse@xxxxxxxxx] On Behalf Of Laine Stump
Sent: Wednesday, March 30, 2016 9:25 PM
To: Libvirt <libvir-list@xxxxxxxxxx>
Cc: Moshe Levi <moshele@xxxxxxxxxxxx>
Subject: Re: [libvirt] Host device assignment driver name vfio/ kvm

 

On 03/29/2016 07:45 AM, Moshe Levi wrote:

Hi,

 

I was  testing Host device assignment  in OpenStack environment where the driver name is vfio or kvm.

 

My setup is as follow:

1.       Fedora 21

2.       Libvirt 1.3.0 which I compiled

3.       OpenStack master

 

I have also other setups with older Libvirt version and the same OpenStack environment.

 

I notice that on my fedora environment the driver name is vfio were in my old environment the driver name is kvm.

 

According to Libvirt documentation default is "vfio" on systems where the VFIO driver is available and loaded, see [1]

I remove the vfio modules by removing vfio, vfio_iommu_type1, vfio_pci but when I boot a vm the drive name is vfio

How can change the driver name to be kvm?

libvirt tries very hard to use vfio rather than legacy kvm, because legacy kvm is old, deprecated, and "declared bad" :-). But it won't changed it to vfio if you've explicitly said that you want to use kvm. If you really want to use legacy kvm device assignment, manually set that in the config. When you do that, if the system you're running on doesn't support it, it will error out rather than switching.

 

 

 

Another thing that I encounter is an error when suspending VM (in OpenStack environment)  when the driver name is  vfio.

In such case I am getting  the following error from Libvirt:

2016-03-28 11:42:59.527 1966 ERROR oslo_messaging.rpc.dispatcher   File "/usr/lib64/python2.7/site-packages/libvirt.py", line 560, in attachDeviceFlags

2016-03-28 11:42:59.527 1966 ERROR oslo_messaging.rpc.dispatcher     if ret == -1: raise libvirtError ('virDomainAttachDeviceFlags() failed', dom=self)

2016-03-28 11:42:59.527 1966 ERROR oslo_messaging.rpc.dispatcher libvirtError: internal error: unable to execute QEMU command 'device_add': Device initialization failed.

 

I would appreciate for some pointers on what can cause this issue.

Assuming that openstack uses libvirt's virDomainSave API I would expect suspending a guest to fail if it had an assigned device (since libvirt implements this by "migrating to disk", and qemu doesn't allow migration of a guest with an assigned device. But your problem is that it's trying to *attach* a device, which I wouldn't consider to be a part of a save or suspend or whatever operation. Is it possible to get more information about what leads up to this?

 

[1] https://libvirt.org/formatdomain.html#elementsHostDev

 

 

 

--

libvir-list mailing list

libvir-list@xxxxxxxxxx

https://www.redhat.com/mailman/listinfo/libvir-list

 

 

 

Attachment: vm.xml
Description: vm.xml

--
libvir-list mailing list
libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list