Bug 1408812

Summary:	[RFE] Implement hotplug for pxb devices
Product:	Red Hat Enterprise Linux Advanced Virtualization	Reporter:	Marcel Apfelbaum <marcel>
Component:	libvirt	Assignee:	Laine Stump <laine>
Status:	CLOSED WONTFIX	QA Contact:	Lili Zhu <lizhu>
Severity:	unspecified	Docs Contact:
Priority:	medium
Version:	8.0	CC:	ailan, chayang, chhu, dyuan, hhan, jherrman, jinzhao, juzhang, laine, marcel, mtessun, virt-maint, xuzhang, yalzhang
Target Milestone:	rc	Keywords:	FutureFeature, TestOnly
Target Release:	---
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	Enhancement
Doc Text:		Story Points:	---
Clone Of:	1361059	Environment:
Last Closed:	2021-01-15 07:29:34 UTC	Type:	Feature Request
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:	1361059, 1529494
Bug Blocks:

Description Marcel Apfelbaum 2016-12-27 12:12:41 UTC

+++ This bug was initially created as a clone of Bug #1361059 +++

Description of problem:
If I hot-plug device into pxb by using libvirt, it reports 'device attach successfully'. However, the device can not be detected unless rebooting guest. So qemu does not support hotplug for pxb yet. 

Version-Release number of selected component (if applicable):
qemu-kvm-rhev-2.6.0-15.el7.x86_64


Libvirt should expose the controllers as hot-pluggable.

Comment 2 Laine Stump 2017-01-03 19:55:40 UTC

Note that in order to do this in an intelligent fashion, we will need Eduardo's qemu patches which report more details about the capabilities of each model of PCI controller (otherwise we would have to allow hotplug for pxb on *all* versions of qemu, which would not end well.)

Comment 3 Laine Stump 2017-03-28 14:48:47 UTC

Actually looking back over this - libvirt already has pci-expander-bus marked as hotpluggable (i.e. we allow hotplugging a device into it), and the way to hotplug into a pcie-expander-bus is to have a pcie-root-port already plugged in, and hotplug to that. In both cases libvirt doesn't complain.

So I don't think there's anything for libvirt to do here except test to be sure it works, correct?

(from what I can see, this is supposed to work already for pcie-expander-bus (Bug 1377160), but has been deferred to 7.5 for pci-expander-bus (Bug 1361059).

So should this just be marked as TestOnly?

(The only downside is that when the host's qemu *doesn't* support hotplug properly, there will be no error indication, but the only thing libvirt could do about that would be a blanket ban for all versions of qemu.)

Comment 4 Marcel Apfelbaum 2017-03-29 14:25:09 UTC

(In reply to Laine Stump from comment #3)
> Actually looking back over this - libvirt already has pci-expander-bus
> marked as hotpluggable (i.e. we allow hotplugging a device into it), and the
> way to hotplug into a pcie-expander-bus is to have a pcie-root-port already
> plugged in, and hotplug to that. In both cases libvirt doesn't complain.
> 
> So I don't think there's anything for libvirt to do here except test to be
> sure it works, correct?
> 

Hi Laine,

If you already "announced" that pxb-pcie supports hotplug, you only have to be sure it actually works.
The downside would be we have a bug before, the pxb-pcie did not support hotplug until now...


> (from what I can see, this is supposed to work already for pcie-expander-bus
> (Bug 1377160), but has been deferred to 7.5 for pci-expander-bus (Bug
> 1361059).
> 
> So should this just be marked as TestOnly?

If is "marked" hotpluggable and libvirt allows already the operation, sure.

Thanks,
Marcel

> 
> (The only downside is that when the host's qemu *doesn't* support hotplug
> properly, there will be no error indication, but the only thing libvirt
> could do about that would be a blanket ban for all versions of qemu.)

Comment 5 Laine Stump 2017-05-03 14:35:08 UTC

The qemu BZ was moved to 7.5, so move this to 7.5 (and mark as TestOnly, since there's nothing for libvirt to do)

Comment 12 RHEL Program Management 2021-01-15 07:29:34 UTC

After evaluating this issue, there are no plans to address it further or fix it in an upcoming release.  Therefore, it is being closed.  If plans change such that this issue will be fixed in an upcoming release, then the bug can be reopened.