On Fri, Feb 23, 2007 at 10:56:04AM +0900, Saori Fukuta wrote: > On Wed, 21 Feb 2007 14:11:12 +0000 "Daniel P. Berrange"wrote: > > This is no good because it is Xen specific - the code breaks when used > > with the QEMU / KVM backend to libvirt. I think this is actually a > > combination of two bugs: > > > > - libvirt looks in xenstore to find the port number, but if it is > > missing hardcodes 5900 + domid. So there is a race condition where > > if virt-install asks for the XML dump, before the framebuffer daemon > > has started up, it will get the wrong port number. > > > > This is a hardcoded fallback port number is a hangover from old Xen > > 3.0.2 where the port number was fixed & not kept in xenstore at all. > > We need to fix libvirt to only use this code on Xen <= 3.0.3 > > Yes, I think so, too. > I try to think about how to fix this bug. Check out the libvir-list archives - I posted a patch for this last night. > > - Once the above fix in libvirt is done, we can make virt-install do > > a loop like the one you have below, but instead of looking in xenstore > > it will look at the port in the XML - waiting for it to change from '-1' > > to an actual port number > > You are right, I will consider again as you pointed out. Excellant. Regards, Dan. -- |=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 -=| |=- Perl modules: http://search.cpan.org/~danberr/ -=| |=- Projects: http://freshmeat.net/~danielpb/ -=| |=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 -=|
