On 11/07/2014 10:55 AM, Ján Tomko wrote:
We only format cpu model for MODE_CUSTOM in domain status XML, but we always format features if they are present. This is a problem if we have a domain using MODE_HOST_PASSTHROUGH that has been managedsaved, then restored, since it now has a feature list but no model in /var/run/libvirt/qemu.
Isn't the real issue that on "virsh start" after managedsave there are features defined for the cpu? According to documentation we don't allow features if domain is using MODE_HOST_PASSTHROUGH, but we allow to start it from managedsave state with those features which I think is a core issue.
Use UPDATE_CPU even for the status XML to prevent libvirt from losing track of the domain. https://bugzilla.redhat.com/show_bug.cgi?id=1030793 https://bugzilla.redhat.com/show_bug.cgi?id=1151885 --- src/conf/domain_conf.c | 1 + 1 file changed, 1 insertion(+) diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index 21309b0..acfb04b 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -19625,6 +19625,7 @@ virDomainSaveStatus(virDomainXMLOptionPtr xmlopt, virDomainObjPtr obj) { unsigned int flags = (VIR_DOMAIN_XML_SECURE | + VIR_DOMAIN_XML_UPDATE_CPU | VIR_DOMAIN_XML_INTERNAL_STATUS | VIR_DOMAIN_XML_INTERNAL_ACTUAL_NET | VIR_DOMAIN_XML_INTERNAL_PCI_ORIG_STATES |
-- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list