Re: [Libvir] PATCH: Specify BIOS clock offset for HVM

On Mon, Jul 16, 2007 at 03:50:07AM -0400, Daniel Veillard wrote: > On Fri, Jul 13, 2007 at 05:07:03PM +0100, Daniel P. Berrange wrote: > > The attached patch introduces a new XML element for specifying information > > about the guest (BIOS) clock. For Xen HVM, and QEMU / KVM guests this is > > used to specifyc whether the guest clock should be in UTC, or localtime. > > The latter is useful for Windows which likes the BIOS to be localtime, > > while the former is useful for all other operating systems. The former is > > of course the default - so no change in existing semantics. > > > > The <clock> element currently takes a single attribute 'offset'. This is > > either 'utc' or 'localtime', but if we port to VMWare, it will also allow > > an arbitrary numeric offset as well as these special constants. > > > > <clock offset='utc'/> > > > > Or > > > > <clock offset='localtime'/> > > > > > > It may later also be desirable to add a 'sync=[yes|no]' attribute to > > specify whether the HV tries to keep the clock in sync with the HV while > > it is running. VMWare has this concept, and so does Xen paravirt - though > > Xen paravirt sets it via /proc/sys/xen/independant_wallclock inside the > > guest and AFAICT doesn't (yet) expose it to the guest config in Dom0. > > > > Anyway, my patch implements use of utc/localtime offsets for Xen, and QEMU > > drivers. I've tested to verify that -localtime gets passed to the QEMU > > process as appropriate. > > Looks fine, the patch also includes a few cleanups. And this should > fix the clock annoyance when running Windows. I assume this will be > automatically added when installing a new Windows guest with virt-manager. Yeah, ignore the unrelated cleanups - I didn't mean to include them - i'll commit them separately. Also ignore the stupid printf() debugging!