Bug 499576

Summary:	RHEL5.4 vt-d: Assigning a PCI device using virt-manager/virt-install fails
Product:	Red Hat Enterprise Linux 5	Reporter:	Mark McLoughlin <markmc>
Component:	python-virtinst	Assignee:	Daniel Berrangé <berrange>
Status:	CLOSED NEXTRELEASE	QA Contact:	Virtualization Bugs <virt-bugs>
Severity:	medium	Docs Contact:
Priority:	high
Version:	5.4	CC:	berrange, crobinso, xen-maint
Target Milestone:	rc
Target Release:	---
Hardware:	All
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:	499267	Environment:
Last Closed:	2009-06-08 15:30:09 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Mark McLoughlin 2009-05-07 09:44:06 UTC

+++ This bug was initially created as a clone of Bug #499267 +++

Just tested virt-manager's device assignment with the following steps:

  * Run virt-manager, open an existing guest and go to the details tab
  * Click "Add hardware", choose "Physical host device" and click "Forward"
  * Choose the appropriate device from the drop down list (e.g. "00:19.0 Interface eth0 (82566DM-2 Gigabit Network Connection)")
  * Click "Forward" and "Finish" 

and it failed with:

Traceback (most recent call last):
  File "/usr/share/virt-manager/virtManager/addhardware.py", line 571, in finish
    self.add_hostdev()
  File "/usr/share/virt-manager/virtManager/addhardware.py", line 607, in add_hostdev
    self._dev.setup()
  File "/usr/lib/python2.6/site-packages/virtinst/VirtualHostDevice.py", line 207, in setup
    raise RuntimeError(_("Could not detach PCI device: %s" % str(e)))
RuntimeError: Could not detach PCI device: 'PCIDevice' object has no attribute 'dettach'

--- Additional comment from markmc on 2009-05-05 17:07:50 EDT ---

Just tried with virt-install --host-device and got the same error

--- Additional comment from markmc on 2009-05-06 09:41:36 EDT ---

Okay, the problem turns out to be that VirtualHostDevice._nodedev is a NodeDevice, not a virNodeDevice yet we do this:

  self._nodedev.dettach()
  self._nodedev.reset()

We need a comprehensive fix which doesn't throw away the virNodeDevice in NodeDeviceParser.lookupNodeName(), but I think we should temporarily disable this dettach/reset code for F11 GA and the test day tomorrow

--- Additional comment from crobinso on 2009-05-06 10:09:36 EDT ---

Created an attachment (id=342658)
Call detach/reset on the correct object

Seems to have done the job, since choosing an arbitrary pci entry in 'virsh nodedev-list' and passing to --host-device managed to lock my machine up :)

--- Additional comment from markmc on 2009-05-06 11:12:13 EDT ---

Looks good to me

--- Additional comment from crobinso on 2009-05-06 12:36:36 EDT ---

Built into python-virtinst-0.400.3-8.fc11.

--- Additional comment from markmc on 2009-05-06 16:21:38 EDT ---

Tag request: https://fedorahosted.org/rel-eng/ticket/1749

* Wed May 06 2009 Cole Robinson <crobinso> - 0.400.3-8.fc11
- Fix PCI assignment (bz 499267)

--- Additional comment from markmc on 2009-05-06 17:37:25 EDT ---

Tagged now

Comment 1 Cole Robinson 2009-05-28 17:50:15 UTC

The patch for this is upstream here:

http://hg.et.redhat.com/cgi-bin/hg-virt.cgi/applications/virtinst--devel/rev/ab3111177ec1

Fix is very low risk, and without it, PCI device assignment doesn't work via virtinst + virt-manager. Since PCI assignment support in the tools is one of the big features for 5.4, we should probably get this in.

Comment 2 Cole Robinson 2009-06-08 15:30:09 UTC

Fixed in python-virtinst-0.400.3-2.el5. Closing as NEXTRELEASE, since this bug (and the affected functionality) has not existed in any released package as it was brought in by the rebase.