pci passthrough fail on fedora 15/kvm/qemu/virt-manager hosting win 7 (all 64 bit)

pci passthrough fail on fedora 15/kvm/qemu/virt-manager hosting win 7 (all 64 bit)

[Erik Flister]

i posted this to virt-tools but got no replies, hoping you guys can help me.  :)


>----- Forwarded Message -----
>>From: Erik Flister <e_flister@yahoo.com>
>>To: erik flister <erik.flister@gmail.com>; "virt-tools-list@redhat.com" <virt-tools-list@redhat.com>
>>Sent: Tuesday, September 20, 2011 9:36 AM
>>Subject: Re: [virt-tools-list] pci passthrough fail on fedora 15/kvm/qemu/virt-manager hosting win 7 (all 64 bit)
>>
>>
>>i found this discussion of the same error:
>>
>>http://www.linux-kvm.com/ content/pci-passthrough-error
>>> "I've solved my problem by deleting the '.save' file in
>>> /var/lib/libvirt/qemu/save"
>>
>>but i don't have that file:
>>
>>> sudo ls -al
/var/lib/libvirt/qemu/save
>>total 8
>>drwxr-xr-x 2 qemu qemu 4096 Sep 16 03:22 .
>>drwxr-x--- 5 qemu qemu 4096 Sep 19 10:13 ..
>>
>>
>>-erik
>>
>>
>>
>>>________________________________
>>>From: erik flister <erik.flister@gmail.com>
>>>To: virt-tools-list@redhat.com
>>>Sent: Monday, September 19, 2011 9:19 PM
>>>Subject: [virt-tools-list] pci passthrough fail on fedora 15/kvm/qemu/virt-manager hosting win 7 (all 64 bit)
>>>
>>>AMD
phenom II X6 1075T proc
>>>ASUS M4A87TD mobo
>>>BIOS ver 2001 (built 3/8/11)
>>>
>>>advanced/cpu config/secure virtual machine mode enabled in BIOS
>>>
>>>not sure if this turns on iommu, how do i verify?
>>>
>>>/proc/cpuinfo has hits for svm but not iommu or vmx.
>>>
>>>> dmesg | grep -i iommu
>>>[    0.000000] Please enable the IOMMU option in the BIOS setup
>>>[    1.515596] PCI-DMA: using GART IOMMU.
>>>[    1.515599] PCI-DMA: Reserving 64MB of IOMMU area in the AGP aperture
>>>
>>>note my pci cards have windows drivers, but not linux drivers as far as i know.
>>>
>>>i have latest yummed kvm/qemu/virt-manager.
>>>
>>>i am using this virtualization method:
>>>http://www.techotopia.com/index.php/Running_Windows_on_Fedora_Using_KVM_Virtualization
>>>
>>>i didn't know to disconnect the pci
cards from the
host before adding
>>>them to the guest until finding:
>>>http://docs.fedoraproject.org/en-US/Fedora/13/html/Virtualization_Guide/sect-Virtualization-PCding_a_PCI_device_to_a_host.html
>>>
>>>> lspci
>>>
>>>02:06.0 Unassigned class [ff00]: National Instruments PCI-6110
>>>02:07.0 Serial controller: NetMos Technology PCI 9865 Multi-I/O Controller
>>>02:07.1 Serial controller: NetMos Technology PCI 9865 Multi-I/O Controller
>>>02:07.2 Parallel controller: NetMos Technology PCI 9865 Multi-I/O Controller
>>>
>>>> lspci -n
>>>
>>>02:06.0 ff00: 1093:14e0
>>>02:07.0 0700: 9710:9865
>>>02:07.1 0700: 9710:9865
>>>02:07.2 0701: 9710:9865
>>>
>>>> virsh nodedev-list | grep pci
>>>
>>>pci_0000_02_06_0
>>>pci_0000_02_07_0
>>>pci_0000_02_07_1
>>>pci_0000_02_07_2
>>>
>>>> sudo virsh
nodedev-dettach
pci_0000_02_06_0
>>>Device pci_0000_02_06_0 dettached
>>>
>>>
>>>but then, after adding it to the guest and trying to boot it, i get:
>>>
>>>Error starting domain: this function is not supported by the
>>>connection driver: Unable to reset PCI device 0000:02:06.0: this
>>>function is not supported by the connection driver: Active
>>>0000:02:07.0 devices on bus with 0000:02:06.0, not doing bus reset
>>>
>>>Traceback (most recent call last):
>>>  File "/usr/share/virt-manager/virtManager/asyncjob.py", line 45, in cb_wrapper
>>>    callback(asyncjob, *args, **kwargs)
>>>  File "/usr/share/virt-manager/virtManager/engine.py", line 959, in asyncfunc
>>>    vm.startup()
>>>  File "/usr/share/virt-manager/virtManager/domain.py", line 1128, in startup
>>>    self._backend.create()
>>>  File "/usr/lib64/python2.7/site-packages/libvirt.py", line 330, in create
>>>    if ret == -1: raise libvirtError
('virDomainCreate() failed', dom=self)
>>>
>>>
>>>
>>>if i disable the other devices on bus 02 and add them to the guest:
>>>> sudo virsh nodedev-dettach pci_0000_02_07_0
>>>Device pci_0000_02_07_0 dettached
>>>
>>>> sudo virsh nodedev-dettach pci_0000_02_07_1
>>>Device pci_0000_02_07_1 dettached
>>>
>>>> sudo virsh nodedev-dettach pci_0000_02_07_2
>>>Device pci_0000_02_07_2 dettached
>>>
>>>i then get this error on booting the guest (but with the same stack
>>>trace as above):
>>>
>>>Error starting domain: cannot send monitor command
>>>'{"execute":"qmp_capabilities"}': Connection reset by peer
>>>
>>>and then networking fails on the host until i reboot, and fedora won't
>>>even shutdown all the way (i have to do a hard reboot by holding down
>>>the power button).
>>>
>>>any idea how i can fix this?  why isn't detaching the device
>>>automatically done when it is added to the guest?  even if i get this
>>>working, it seems i will
have to detach stuff every time i reboot the
>>>host?  finally, in this case, i actually do want to give everything on
>>>bus 02 to the guest, but what if i didn't?
>>>
>>>thanks for your help!
>>>-erik
>>>
>>>
>>>see also iommu/bios stuff here:
>>>http://docs.fedoraproject.org/en-US/Fedora/13/html-single/Virtualization_Guide/index.html
>>>
>>>_______________________________________________
>>>virt-tools-list mailing list
>>>virt-tools-list@redhat.com
>>>https://www.redhat.com/mailman/listinfo/virt-tools-list

Re: pci passthrough fail on fedora 15/kvm/qemu/virt-manager hosting win 7 (all 64 bit)

[Alex Williamson]

On Wed, 2011-09-21 at 10:23 -0700, Erik Flister wrote:
> i posted this to virt-tools but got no replies, hoping you guys can help me.  :)
> 

> >>>AMD
> phenom II X6 1075T proc
> >>>ASUS M4A87TD mobo

Google says this is an AMD 870 chipset, so you do not have AMD IOMMU
(AMD-Vi) support :(

> >>>> lspci
> >>>
> >>>02:06.0 Unassigned class [ff00]: National Instruments PCI-6110
> >>>02:07.0 Serial controller: NetMos Technology PCI 9865 Multi-I/O Controller
> >>>02:07.1 Serial controller: NetMos Technology PCI 9865 Multi-I/O Controller
> >>>02:07.2 Parallel controller: NetMos Technology PCI 9865 Multi-I/O Controller
> >>>
> >>>> lspci -n
> >>>
> >>>02:06.0 ff00: 1093:14e0
> >>>02:07.0 0700: 9710:9865
> >>>02:07.1 0700: 9710:9865
> >>>02:07.2 0701: 9710:9865
> >>>
> >>>> virsh nodedev-list | grep pci
> >>>
> >>>pci_0000_02_06_0
> >>>pci_0000_02_07_0
> >>>pci_0000_02_07_1
> >>>pci_0000_02_07_2
> >>>
> >>>> sudo virsh
> nodedev-dettach
> pci_0000_02_06_0
> >>>Device pci_0000_02_06_0 dettached
> >>>
> >>>
> >>>but then, after adding it to the guest and trying to boot it, i get:
> >>>
> >>>Error starting domain: this function is not supported by the
> >>>connection driver: Unable to reset PCI device 0000:02:06.0: this
> >>>function is not supported by the connection driver: Active
> >>>0000:02:07.0 devices on bus with 0000:02:06.0, not doing bus reset

This is probably telling you that the device you want to assign doesn't
support a reset mechanism (FLR or D3hot), so it wants to do a secondary
bus reset, but there are other devices on the same bus.

> >>>if i disable the other devices on bus 02 and add them to the guest:
> >>>> sudo virsh nodedev-dettach pci_0000_02_07_0
> >>>Device pci_0000_02_07_0 dettached
> >>>
> >>>> sudo virsh nodedev-dettach pci_0000_02_07_1
> >>>Device pci_0000_02_07_1 dettached
> >>>
> >>>> sudo virsh nodedev-dettach pci_0000_02_07_2
> >>>Device pci_0000_02_07_2 dettached
> >>>
> >>>i then get this error on booting the guest (but with the same stack
> >>>trace as above):
> >>>
> >>>Error starting domain: cannot send monitor command
> >>>'{"execute":"qmp_capabilities"}': Connection reset by peer
> >>>
> >>>and then networking fails on the host until i reboot, and fedora won't
> >>>even shutdown all the way (i have to do a hard reboot by holding down
> >>>the power button).
> >>>
> >>>any idea how i can fix this?  why isn't detaching the device
> >>>automatically done when it is added to the guest?  even if i get this
> >>>working, it seems i will
> have to detach stuff every time i reboot the
> >>>host?  finally, in this case, i actually do want to give everything on
> >>>bus 02 to the guest, but what if i didn't?

Getting it working seems like it's going to require new hardware since
you don't have a device assignment capable IOMMU.  Thanks,

Alex