Re: [Qemu-discuss] Error starting domain: vfio-pci

[Bandan Das]

Gabriel Laupre <address@hidden> writes:

>> Yes, indeed it is. What distro is this ? Do you have SELinux or any other
> security feature enabled ? Can you please verify that the file has a
> appropriate label if SELinux is enabled ? (ls -lZ /dev/vfio/vfio)
> My distrib:
> address@hidden ~]# cat /proc/version
> Linux version 3.10.0-229.1.2.el7.x86_64 (address@hidden)
> (gcc version 4.8.2 20140120 (Red Hat 4.8.2-16) (GCC) ) #1 SMP Fri Mar 27
> 03:04:26 UTC 2015
> address@hidden ~]# cat /etc/centos-release
> CentOS Linux release 7.1.1503 (Core)
>
> address@hidden ~]# ls -lZ /dev/vfio/vfio
> crw-rw-rw- root root ?                                /dev/vfio/vfio
>
> SELinux is disabled:
> address@hidden ~]# getenforce
> Disabled
>
> I guess no other security feature is enabled that I am aware of. I once had
> a message saying that it can be one of the following issues (listing the
> 5). So I guess it can be any combination of those issues, even something
> completely different.

Ugh, I am out of options! Can you please try a few more things: Can you try
running qemu directly and see if you see the same behavior ? If you still
haven't tried running as root, please try that too. Also, please check dmesg
for any vfio related errors.

> libvirtError: internal error: process exited while connecting to
> monitor: 2015-05-19T21:46:21.935043Z qemu-kvm: -device
> vfio-pci,host=04:10.4,id=hostdev0,bus=pci.0,addr=0x3: vfio: failed to
> open /dev/vfio/vfio: Operation not permitted
Well, this is the first error from vfio_connect_container() when it does:
    fd = qemu_open("/dev/vfio/vfio", O_RDWR);
    if (fd < 0) {
        error_report("vfio: failed to open /dev/vfio/vfio: %m");
        ret = -errno;
...

The rest are followup errors printed from the other functions in the
stack due to this error.

Bandan

> 2015-05-19T21:46:21.935091Z qemu-kvm: -device
> vfio-pci,host=04:10.4,id=hostdev0,bus=pci.0,addr=0x3: vfio: failed to
> setup container for group 24
> 2015-05-19T21:46:21.935107Z qemu-kvm: -device
> vfio-pci,host=04:10.4,id=hostdev0,bus=pci.0,addr=0x3: vfio: failed to
> get group 24
> 2015-05-19T21:46:21.935135Z qemu-kvm: -device
> vfio-pci,host=04:10.4,id=hostdev0,bus=pci.0,addr=0x3: Device
> initialization failed.
> 2015-05-19T21:46:21.935157Z qemu-kvm: -device
> vfio-pci,host=04:10.4,id=hostdev0,bus=pci.0,addr=0x3: Device
> 'vfio-pci' could not be initialized
>
>
>
>
> 2015-05-19 21:17 GMT-08:00 Bandan Das <address@hidden>:
>
>>
>> > On May 20, 2015, at 12:29 AM, Gabriel Laupre <address@hidden> wrote:
>> >
>> > Thank Bandan,
>> >
>> >
>> > > Assuming you are on an intel box, have you booted your kernel with
>> intel_iommu=on ?
>> > Yes, I have booted my kernel with the intel_iommu=on. (I don't remember
>> how to check that now though ^^)
>> >
>> > > Please paste the output of dmesg | grep -e DMAR -e IOMMU ?
>> > address@hidden ~]# dmesg | grep -e DMAR -e IOMMU
>> > [    0.000000] ACPI: DMAR 00000000bf79e0c0 00118 (v01    AMI  OEMDMAR
>> 00000001 MSFT 00000097)
>> > [    0.000000] Intel-IOMMU: enabled
>> > [    0.039149] dmar: IOMMU 0: reg_base_addr fbffe000 ver 1:0 cap
>> c90780106f0462 ecap f020f6
>> > [    0.550126] IOMMU 0 0xfbffe000: using Queued invalidation
>> > [    0.550131] IOMMU: Setting RMRR:
>> > [    0.550149] IOMMU: Setting identity map for device 0000:00:1a.0
>> [0xbf7ec000 - 0xbf7fffff]
>> > [    0.550184] IOMMU: Setting identity map for device 0000:00:1a.1
>> [0xbf7ec000 - 0xbf7fffff]
>> > [    0.550211] IOMMU: Setting identity map for device 0000:00:1a.2
>> [0xbf7ec000 - 0xbf7fffff]
>> > [    0.550241] IOMMU: Setting identity map for device 0000:00:1a.7
>> [0xbf7ec000 - 0xbf7fffff]
>> > [    0.550272] IOMMU: Setting identity map for device 0000:00:1d.0
>> [0xbf7ec000 - 0xbf7fffff]
>> > [    0.550302] IOMMU: Setting identity map for device 0000:00:1d.1
>> [0xbf7ec000 - 0xbf7fffff]
>> > [    0.550329] IOMMU: Setting identity map for device 0000:00:1d.2
>> [0xbf7ec000 - 0xbf7fffff]
>> > [    0.550358] IOMMU: Setting identity map for device 0000:00:1d.7
>> [0xbf7ec000 - 0xbf7fffff]
>> > [    0.550375] IOMMU: Setting identity map for device 0000:00:1a.0
>> [0xec000 - 0xeffff]
>> > [    0.550387] IOMMU: Setting identity map for device 0000:00:1a.1
>> [0xec000 - 0xeffff]
>> > [    0.550399] IOMMU: Setting identity map for device 0000:00:1a.2
>> [0xec000 - 0xeffff]
>> > [    0.550410] IOMMU: Setting identity map for device 0000:00:1a.7
>> [0xec000 - 0xeffff]
>> > [    0.550421] IOMMU: Setting identity map for device 0000:00:1d.0
>> [0xec000 - 0xeffff]
>> > [    0.550433] IOMMU: Setting identity map for device 0000:00:1d.1
>> [0xec000 - 0xeffff]
>> > [    0.550444] IOMMU: Setting identity map for device 0000:00:1d.2
>> [0xec000 - 0xeffff]
>> > [    0.550458] IOMMU: Setting identity map for device 0000:00:1d.7
>> [0xec000 - 0xeffff]
>> > [    0.550471] IOMMU: Prepare 0-16MiB unity mapping for LPC
>> > [    0.550483] IOMMU: Setting identity map for device 0000:00:1f.0 [0x0
>> - 0xffffff]
>> >
>>
>> Yeah, this looks ok. Actually, taking a second look, I can’t think of
>> anyway how this could be related to file permissions on /dev/vfio/vfio.
>>
>> > > Why does opening /dev/vfio/vfio fail ? Can you please confirm that you
>> have read/write permissions as the user you are trying to run ?
>> > address@hidden ~]# cd /dev/vfio/
>> > address@hidden vfio]# ls -la | grep vfio
>> > crw-rw-rw-  1 root root 10, 196 May 18 11:54 vfio
>> > The right should be okay I guess.
>> >
>> Yes, indeed it is. What distro is this ? Do you have SELinux or any other
>> security feature enabled ? Can you please verify that the file has a
>> appropriate label if SELinux is enabled ? (ls -lZ /dev/vfio/vfio)
>>
>> Bandan
>> >
>> >
>> >
>> > 2015-05-19 18:54 GMT-08:00 Bandan Das <address@hidden>:
>> >
>> > Hello Gabriel,
>> >
>> > > On May 19, 2015, at 8:03 PM, Gabriel Laupre <address@hidden> wrote:
>> > >
>> > > Hello everyone,
>> > >
>> > > I am using a Centos 7.1 machine with the kernel 3.10.229. I want to
>> use my host with SR-IOV to use a virtual function on my NIC as the vNIC in
>> my new VM.
>> > >
>> > > I have an instance started with a old NIC using macvtap that I want to
>> change. I am using the
>> > > virsh edit instance-00000034
>> > > command to edit the XML configuration to add the new device I want to
>> attach.
>> > …
>> > Assuming you are on an intel box, have you booted your kernel with
>> intel_iommu=on ?
>> > Please paste the output of dmesg | grep -e DMAR -e IOMMU ?
>> >
>> > > When I try to reboot the VM I get this error:
>> > > Error starting domain: internal error: process exited while connecting
>> to monitor: 2015-05-19T21:46:21.935043Z qemu-kvm: -device
>> vfio-pci,host=04:10.4,id=hostdev0,bus=pci.0,addr=0x3: vfio: failed to open
>> /dev/vfio/vfio: Operation not permitted
>> > Why does opening /dev/vfio/vfio fail ? Can you please confirm that you
>> have read/write permissions as the user you are trying to run ?
>> >
>> > > 2015-05-19T21:46:21.935091Z qemu-kvm: -device
>> vfio-pci,host=04:10.4,id=hostdev0,bus=pci.0,addr=0x3: vfio: failed to setup
>> container for group 24
>> > > 2015-05-19T21:46:21.935107Z qemu-kvm: -device
>> vfio-pci,host=04:10.4,id=hostdev0,bus=pci.0,addr=0x3: vfio: failed to get
>> group 24
>> > > 2015-05-19T21:46:21.935135Z qemu-kvm: -device
>> vfio-pci,host=04:10.4,id=hostdev0,bus=pci.0,addr=0x3: Device initialization
>> failed.
>> > > 2015-05-19T21:46:21.935157Z qemu-kvm: -device
>> vfio-pci,host=04:10.4,id=hostdev0,bus=pci.0,addr=0x3: Device 'vfio-pci'
>> could not be initialized
>> > >
>> > > total Trace here: http://sprunge.us/XZFB
>> > >
>> > > Any idea how to fix that?
>> > >
>> > > Thank you very much :)
>> > >
>> > > Gabriel
>> > >
>> >
>> >
>>
>>