[RFC PATCH] Behaviour of topology, vcpus and maxcpus

[Nikunj A Dadhania <nikunj@xxxxxxxxxxxxxxxxxx>]

I have observed the following with libvirt xml:

<vcpu>6</vcpu>
<cpu>
  <topology sockets='1' cores='4' threads='2'/>
</cpu>

So according to the topology maximum supported is 8 vcpus, while the
libvirt sets that to 6 - specified in <vcpu> tag. Shouldn't libvirt
error this out as the mismatch between the topology definition and the
vcpu count?

There is an upper bound check already in place where if we have maxcpu >
topology-supported cpus, that errors out. For eg.

<vcpu>12</vcpu>
<cpu>
  <topology sockets='1' cores='4' threads='2'/>
</cpu>

The below patch make sures that libvirt does the lower bound check as
well.

Regards
Nikunj


diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c
index f96110b..aa25940 100644
--- a/src/conf/domain_conf.c
+++ b/src/conf/domain_conf.c
@@ -12933,6 +12933,14 @@ virDomainDefParseXML(xmlDocPtr xml,
             goto error;
         }
 
+        if (def->cpu->sockets &&
+            def->maxvcpus <
+            def->cpu->sockets * def->cpu->cores * def->cpu->threads) {
+            virReportError(VIR_ERR_XML_DETAIL, "%s",
+                           _("CPUs provided is less than topology"));
+            goto error;
+        }
+
         if (def->cpu->cells_cpus > def->maxvcpus) {
             virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                            _("Number of CPUs in <numa> exceeds the"

--
libvir-list mailing list
libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list