[Openstack] [openstack-dev] PCI pass-through SRIOV

Kamsali, RaghavendraChari (Artesyn) RaghavendraChari.Kamsali at artesyn.com
Wed May 27 06:09:42 UTC 2015


Hi,

I resolved it by adding or uncommenting  "/dev/vfio/vfio" in cgroup_device_acl list in /etc/libvirtd/qemu.conf file and restart libvird service .

Now when I listed nova instances am not getting any Network shown 

[stack at Controller images]$ nova list
+--------------------------------------+----------+--------+------------+-------------+----------+
| ID                                   | Name     | Status | Task State | Power State | Networks |
+--------------------------------------+----------+--------+------------+-------------+----------+
| 6cd81ab3-d60f-4511-ad5d-7d337af3cacd | ubuntuvm | ACTIVE | -          | Running     |          |
+--------------------------------------+----------+--------+------------+-------------+----------+
[stack at Controller images]$



-----Original Message-----
From: Steve Gordon [mailto:sgordon at redhat.com] 
Sent: Tuesday, May 26, 2015 9:45 PM
To: Moshe Levi
Cc: Kamsali, RaghavendraChari [ENGINEERING/IN]; OpenStack Development Mailing List (not for usage questions); openstack at lists.openstack.org
Subject: Re: [Openstack] [openstack-dev] PCI pass-through SRIOV

----- Original Message -----
> From: "Moshe Levi" <moshele at mellanox.com>
> To: "RaghavendraChari Kamsali (Artesyn)" 
> <RaghavendraChari.Kamsali at artesyn.com>, "OpenStack Development Mailing 
> List
> 
> This is a different  error
> 
> 2015-05-26 13:34:08.081 TRACE nova.compute.manager [instance:
> 101776a0-cd2e-47b9-bdc4-1097782201c6]     if ret == -1: raise libvirtError
> ('virDomainCreateWithFlags() failed', dom=self)
> 2015-05-26 13:34:08.081 TRACE nova.compute.manager [instance:
> 101776a0-cd2e-47b9-bdc4-1097782201c6] libvirtError: internal error: 
> process exited while connecting to monitor: 2015-05-26T04:34:07.980897Z qemu-kvm:
> -device vfio-pci,host=81:02.3,id=hostdev0,bus=pci.0,addr=0x4: vfio: 
> failed to open /dev/vfio/vfio: Operation not permitted
> 2015-05-26 13:34:08.081 TRACE nova.compute.manager [instance:
> 101776a0-cd2e-47b9-bdc4-1097782201c6] 2015-05-26T04:34:07.980951Z qemu-kvm:
> -device vfio-pci,host=81:02.3,id=hostdev0,bus=pci.0,addr=0x4: vfio: 
> failed to setup container for group 49
> 2015-05-26 13:34:08.081 TRACE nova.compute.manager [instance:
> 101776a0-cd2e-47b9-bdc4-1097782201c6] 2015-05-26T04:34:07.980970Z qemu-kvm:
> -device vfio-pci,host=81:02.3,id=hostdev0,bus=pci.0,addr=0x4: vfio: 
> failed to get group 49
> 2015-05-26 13:34:08.081 TRACE nova.compute.manager [instance:
> 101776a0-cd2e-47b9-bdc4-1097782201c6] 2015-05-26T04:34:07.980995Z qemu-kvm:
> -device vfio-pci,host=81:02.3,id=hostdev0,bus=pci.0,addr=0x4: Device 
> initialization failed.
> 2015-05-26 13:34:08.081 TRACE nova.compute.manager [instance:
> 101776a0-cd2e-47b9-bdc4-1097782201c6] 2015-05-26T04:34:07.981019Z qemu-kvm:
> -device vfio-pci,host=81:02.3,id=hostdev0,bus=pci.0,addr=0x4: Device 
> 'vfio-pci' could not be initialized
> 
> You are using intel card therefore I think you should contact them and 
> ask if this card is supported.

In addition there are a number of Intel cards for which ACS quirks had to be added to the kernel, this was done fairly recently in patches like this one:

    http://www.spinics.net/lists/kernel/msg1951202.html

You may want to check whether a) your card is one of those impacted and b) your kernel has these patches, though your output does not appear to be an exact match for what we were seeing in https://bugzilla.redhat.com/show_bug.cgi?id=1141399

Thanks,

--
Steve Gordon, RHCE
Sr. Technical Product Manager,
Red Hat Enterprise Linux OpenStack Platform


More information about the Openstack mailing list