[libvirt] PATCH: 11/14: Change XM driver in-memory objects

[Daniel Veillard]

On Thu, Jul 24, 2008 at 05:10:55PM +0100, Daniel P. Berrange wrote:
> On Thu, Jul 24, 2008 at 10:44:18AM -0400, Daniel Veillard wrote:
> > On Tue, Jul 08, 2008 at 05:40:17PM +0100, Daniel P. Berrange wrote:
> > > This is a refactoring of the XM driver. Previously we would store
> > > the virConfPtr objects as our master 'in memory' representation
> > > of inactive domains. This switch it over to using virDomainDefPtr
> > > objects instead. The code for reading/writing the config files is
> > > unchanged at this time.
> > 
> > [...]
> > > @@ -1291,9 +1217,10 @@ int xenXMDomainPinVcpu(virDomainPtr doma
> > >      xenXMConfCachePtr entry;
> > >      virBuffer mapbuf = VIR_BUFFER_INITIALIZER;
> > >      char *mapstr = NULL;
> > > -    char *ranges = NULL;
> > >      int i, j, n, comma = 0;
> > >      int ret = -1;
> > > +    char *cpuset = NULL;
> > > +    int maxcpu = 4096;
> > 
> >   hum, we use MAX_VIRT_CPUS at places
> 
> Yeah, we need todo better here. The problem is Xen's idea of
> max CPU's is different from Linux's :-)

  maybe a LIBVIRT_MAX_CPU exported from the libvirt.h API so that
users know what is the limitation we use internally (or make everything
dynamic but it's maybe too much).

> > > +++ b/tests/xmconfigdata/test-fullvirt-new-cdrom.xml	Mon Jul 07 10:11:30 2008 -0400
> >  I'm just a bit surprized by that addition, is that derived from 
> > the features set ? I don't see why the arch can't be x86-64 for example
> > just based on the tests/xmconfigdata/test-fullvirt-new-cdrom.cfg config
> > data.
> 
> What's happening is that in the Xen test suite we now define a static
> set of XML capabilities, which are independant of the host machine
> running the test. In the Xen case I defined the test suite to be a
> 32-bit machine, so the test XML files have to use 'i686'. Outside of
> the tst suite, the real capabilities data is fetched from the hypervisor
> so supports whatever is appropriate - i686, x86_64, ia64, ppc, etc.

  okay, makes sense actually we don't want the regression test to break
on every different arch.

  +1 let's push it !

Daniel

-- 
Red Hat Virtualization group http://redhat.com/virtualization/
Daniel Veillard      | virtualization library  http://libvirt.org/
veillard at redhat.com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine  http://rpmfind.net/